What changed in Crawl Foundry
A dated record of product changes, workflow improvements, fixes, and the occasional update that changes existing behavior.
A free starting point for keyword research
The research guide now shows how to build an initial keyword list from customer questions, your own Search Console data and a review of search results. A linked guide explains how to organize the shortlist and includes a CSV import example. The guide explains unknown metrics and optional paid data requests. The English and German AI Mode glossary entries now distinguish Search Console visibility from recorded GA4 visits and link to the traffic checker.
CSV examples and clearer keyword research guides
The keyword-management guide now includes a downloadable CSV example. The AI traffic checker explains which export columns it needs and what to check when no AI source is recognized. A dated example on the SEO MCP page shows how we selected 28 relevant keywords from 791 research suggestions and saved them in separate market lists.
Direct execution for approved MCP keyword-data requests
Keyword-data requests approved through MCP no longer wait for the regular background check to start. The confirmed MCP price also applies when data processing starts. Processing no longer stops with an incorrect duplicate-job error. Unfinished requests keep their existing recovery checks. Price confirmation and spending limits continue to apply.
Download a CSV format example for the Keyword Database
Download a small CSV example from the Keyword Database page and compare its three rows with the import preview. It uses synthetic search terms and the keyword, list and tags columns. No search-volume or difficulty measurements are included. Downloading the file does not start an import or request new data.
Clearer guidance for CSV imports and keyword difficulty
The keyword database page now explains the existing keyword, list and tags CSV columns, the import preview and exports from a selected list more clearly. The English Keyword Difficulty glossary shows why a KD score of 30 is not a 30% chance of ranking.
Launch video on the English and German website
The English and German homepages and About pages now use the matching YouTube uploads for the launch video. Choose to load YouTube when you want to watch, and stop the player at any time.
Keep reading in German
Related article links on German blog pages now open the German versions, so you can continue reading in the same language.
Clear availability notices for MCP and agent API
If the agent service cannot safely accept requests, the MCP and API page shows a temporary availability notice. A manual check lets you try again. Other app features remain accessible. Public MCP sign-in discovery avoids unnecessary Redis commands. Protected agent requests retain their limits and pause during a dependency outage.
Explore analytics directly in your charts
Explore trends in a full-width chart and open daily values when needed. Pie legends filter segments and chart hover highlights matching table rows. Search Console shows four comparisons beside each table. Sitemaps are compact, while index inspection groups repeated notices into expandable details. Search type reports also accept ungrouped responses and keep unavailable positions empty. Analytics detail tabs preserve your scroll position. Select Google trends by clicking anywhere in a metric field, with refresh and connection controls beside the main tabs. Setup and waiting screens guide you before the first measurement. Saved statistics stay visible during service outages, with recovery actions when a refresh fails. Interrupted installations no longer stay busy indefinitely.
Readable analytics charts and more Search Console detail
Read complete domains and URL paths beside ranked charts. Devices, browsers and operating systems use ring and pie charts. Performance trends show quality thresholds alongside their measured values. Search Console statistics now include top queries, pages, countries, devices, search appearances and search types. Explore submitted sitemaps and a sample of URLs with crawl, canonical and structured data details. Failed sections retain their own saved data. New websites can open monitoring or Google setup directly from the standard start banner.
Domain research tools for organization members
The Expired Domain Database, AI Visibility Checker and Backlink Checker are available from the Tools navigation group to members with the relevant permissions. The checkers are free and share an allowance of 50 successful checks per user per day. Failed checks do not consume that allowance, and previous results remain private to the person who checked the domain. In the domain database, viewers can search the full inventory and inspect saved domains. Editors can organize shared lists and request enrichment after confirming its price. Import and infrastructure management remain restricted to administrators. AI Visibility shows pages captured in Common Crawl. The Backlink Checker shows historical references between domains, with additional source information. These observations do not establish AI citations or the exact page containing a backlink.
Compact reports in the admin AI Visibility preview
The admin preview of the AI Visibility Checker now shows captured URLs in a compact, searchable table with 10 or 20 rows per page. Domain checks can retrieve up to 500 URLs and identify incomplete results. Matching pages from an accessible completed Site Audit can open directly in the page view. Recent results are easier to recognize as clickable. The report keeps the domain, date and retrieval status in view; source details and explanations open on demand. The checker remains an admin preview. Common Crawl records historical page captures; these results do not measure AI citations or complete website coverage.
Explore analytics trends and keep saved Google data
Choose a metric in Crawl Foundry Analytics to view its measured daily trend. Countries, devices, sources and pages pair their tables with charts. Website performance values use quality colors. Visited dashboard views retain their data and chart selection while you switch tabs. Hidden views pause automatic provider refreshes. If Google fails or returns no new observations, the last successful data stays visible with its original date and a support code. Ranking summaries show measured keywords and offer the tracking setup when measurements are missing.
More SEO checks and shareable results for browser assistants
Compatible browser assistants can run the AI Robots.txt Checker and check answer structure from a public URL or supplied content. Results open in the same reports you use on the website. URL checks retain the existing verification and request limits. Calculator links restore the usage inputs and selected month. AI Crawler reports can be copied or downloaded as JSON, with measured facts, priorities and links to the relevant checker or documentation. The export omits page text, response headers and URL query strings. The SEO MCP setup guide returns the existing instructions for your chosen assistant and selects its visible setup tab. Tool usage, outcome and duration are measured only with analytics consent. These browser tools use the experimental WebMCP API; the normal website controls remain available.
Public search, crawler checks and contact drafts for browser assistants
Compatible browser assistants can search public product pages, FAQ and documentation, run the AI Crawler Checker and prepare a contact request. Search results and crawler reports appear on the website. Contact drafts stay in the form for you to review and send. The cost calculator now compares the first month with a following month for the same usage. The first month includes one-time trial balance; the following month assumes no unused balance from earlier months. Both estimates show payments before tax. Search and the month comparison are also available through the normal website controls. These tools use the experimental WebMCP browser API. Crawler checks retain the existing verification and rate limits. If documentation search is unavailable, the search shows the available website results and a notice.
Important pages and crawl comparisons in the admin checker
The admin preview of AI Visibility Checker can check selected important pages, group captured URLs by website area and language, and suggest actions for historical errors, redirects and URL variants. Matching Site Audit entries link directly to the page view; an audit starts only through its normal workflow. Small saved domain results provide a baseline for the next crawl. Incomplete samples do not produce misleading missing-page counts. This preview uses Common Crawl captures as evidence and does not measure AI citations or training use.
Referring websites in the admin Backlink Checker
The admin preview of Backlink Checker shows referring domains and hosts from our own Common Crawl graph. Workspace domains are suggested, and a compact table supports search and pagination. The total refers to the graph, with up to 200 source websites displayed. Successful checks share the existing daily allowance with AI Visibility. Failed checks do not consume it. AI Visibility also distinguishes ordinary extensionless pages from files and uses the primary content language. Redirects and failed captures no longer distort the language summary. Graph links are historical evidence. Individual source URLs, anchor text and live verification are not included yet. The report labels the crawl period with month names and identifies website links as homepages. Both checkers use one allowance of 50 successful checks per user per day; the preview remains restricted to admins.
New SEO comparisons, practical workflows and API guidance
The Solutions menu now links to Ahrefs, Semrush and Screaming Frog alternative pages. Each uses the OpenSEO comparison layout, explains where each tool fits, compares workflows and payment models, and links to official product and pricing sources. The comparison does not claim that every feature has a direct replacement. Two new workflow pages explain technical SEO audits and keyword research for agencies, from the initial question to an evidence-based handover. The primary navigation still has five sections. The documentation now includes an API and MCP integration introduction with Swagger, OpenAPI, credential handling, workspace selection and a read-only first request. API keys are managed in the application and are not sent by email. Each comparison answers specific questions about changing tools and comparing project costs. The workflow pages include practical review checklists, explain when a public free tool or a full Site Audit fits, and show how to connect keyword research to existing client pages.
Read chart values and explore your visitors
Dashboard chart tooltips now show the full date and value above the chart. You can reset the organization’s metric selection to the five free visitor statistics before applying it. Open Visitor statistics on a monitored website to see countries, devices, browsers, operating systems, languages, traffic channels, referring websites and pages. Details show readable names, visitor counts and shares from the saved statistics. Each website now shows up to five available metrics in your chosen order. Change the priority in the metric selection. Additional selected metrics replace missing data, and measured zeros remain visible.
Google Search Console and website performance in your dashboard
Switch between your website overview and Google Search Console in the dashboard. Connected websites show clicks, impressions, click-through rate and average position, with a comparison to the previous period. Both views keep the same website headers and saved order. The existing visitor snippet supports real-user performance measurements: LCP, INP, CLS, FCP and TTFB. Choose them in the metric filter when measurements are available. Missing observations stay empty, and saved statistics remain visible when a source is unavailable. Performance measurement now starts automatically through existing snippets, without another tag. New visits supply the measurements. In the position chart, better rankings appear higher while tooltips keep the actual position values. Connect additional websites through a Google account already authorized in your organization. Choose an existing connection and a matching Search Console property without repeating the Google login.
Set up website monitoring from the Dashboard
Start monitoring from your organization’s Dashboard. The Website overview keeps its visitor statistics entry for websites with monitoring enabled. Installation instructions now explain how to replace the previous monitoring script with the combined Crawl Foundry snippet. If the installation check cannot find it, the message identifies the workspace’s snippet that needs to be installed.
Product tours and SEO explanations on the website
Matching product tours and SEO explanations are available on selected feature pages, guides, glossary entries, and blog articles in English and German. Each video has a preview image, captions, native playback controls and a description of what it covers. Videos load when you choose to play them. Product tours use an example workspace and link to Crawl Foundry so you can try the workflow yourself.
Clearer failures and shared headers in the admin Visibility Checker
The admin AI Visibility preview now allows 50 completed checks per day. Technical failures leave the allowance unchanged and are distinguished from an empty crawl result. A request pause shows the actual remaining time. The checker and organization dashboard now share the standard page header with help, problem reporting and a new-analysis action. Common Crawl captures remain historical evidence, not a measure of AI citations. The checker remains available to admins only.
Understand your visitor statistics and refresh them when you need
Website statistics now explain visitors, visits, pageviews, bounce rate and average visit duration. One central button refreshes the connected websites in your organization, including websites on other dashboard pages. While updates are running, the button shows their progress. Each card retains its last successful update time. Visible websites update automatically every five minutes while the dashboard is open. If collection has not recorded any pageviews yet, the dashboard explains what to check. Saved statistics remain visible when an update fails. Websites with visitor statistics appear first, sorted by visitors. Drag the handle to save a custom order for your organization. The visitor chart compares the last 30 days with the previous 30 days. Bounce rate and visit duration gain daily trends when enough measurements are available. The compact detail view groups countries, devices, sources and pages into tabs. The website overview starts with Crawl Foundry Analytics. Separate tabs show the existing SEO overview and Google Search Console. Countries, devices, browsers, sources and pages are available in Analytics. Google statistics can also be opened from the organization dashboard. Better search positions appear higher in the chart. Manual refresh now fetches new performance measurements too. Missing measurements are explicitly identified.
Clearer Brain crawl outcomes and accessible Agent settings
Brain crawls can authorize paid extraction calls correctly and use a current stable extraction model. Failed extractions count as failures, and a crawl that fails on every page no longer appears as successfully completed. Organization owners can open Agent, MCP & API settings with their organization permissions, including organizations created during onboarding. API access lists use the selected organization. Keyword research shows search-volume coverage directly and explains when only some metrics are available. Activity reports distinguish failed usage from failed charges.
Page information and cost estimates for browser assistants
Compatible browser assistants can retrieve a structured overview of selected public product and pricing pages. They can also use the homepage calculator to estimate the first month's costs and display the same result on the page. The integration uses the experimental WebMCP browser API. Estimates include the one-time trial balance and show amounts before tax. The normal website remains available in browsers without WebMCP support.
More room for expired-domain listings and readable table headers
The expired-domain database uses the full available window height. A shorter header and fewer text rows leave more room for listings, while pagination stays visible below the scrolling table. Its table surfaces now follow the keyword database styling. Both databases use opaque table headers so scrolled rows no longer show through the column labels. Range filters cover the full auction inventory, including price, age, auction end and available SEO metrics. Common Crawl presence is available as a separate one-click filter above the table. Stable continuation pages keep browsing consistent while feeds are refreshed. Link-count filters become available when their measurements are complete, and paid SEO enrichment remains a separate action on selected domains.
See your websites together
The organization Dashboard brings your website workspaces together. Its default metrics are free visitor statistics: visitors, visits, pageviews, bounce rate and visit duration. An organization-wide metric selection lets you add existing organic keyword and traffic estimates, referring domains and audit health scores. Search by website or workspace and open a website to switch directly to its Site Overview. The Crawl Foundry logo also returns to the organization Dashboard. The compact header keeps website search close to the overview. Organizations without a website workspace get the same guided start screen as the other tools, with a direct action to create their first website workspace. Websites without measurements stay compact, showing their name and a direct Site Overview link. Websites without tracking show a Start monitoring action that opens snippet setup. Available measurements and saved trends remain visible. Monitoring setup and Site Audit now share the same step-based modal and ownership verification controls. The monitoring assistant guides you through adding and checking one script, while existing DNS, file and meta-tag proofs remain available in Site Audit. Pending checks prevent conflicting actions and workspace changes clear local installation instructions. Installation checks now distinguish an unreachable homepage from unavailable visitor statistics and check the HTML head without downloading a large page body. The former workspace Dashboard is now called Site Overview. Your existing links continue to open the same workspace briefing. Site Audit Portfolio remains available for examining technical audits across websites. Missing measurements stay visibly unavailable, and visitor analytics remains separate from SEO traffic estimates. One website script verifies ownership and collects visitor statistics. The Visitors view includes countries, devices, browsers, pages and traffic sources. Statistics load from saved snapshots, so a tracking service outage keeps website navigation usable and preserves the last successful measurements.
Clearer keyword workflows and glossary explanations
The keyword and Site Audit pages explain what each workflow does, with practical examples and links to the next useful step. Keyword Database and Keyword Management make the distinction between free organization and separately purchased data clearer. The keyword research guide separates finding ideas from choosing a target page. The AI Traffic Checker explains which CSV values it analyzes locally and when an optional URL check sends selected addresses to the server. Glossary articles use their editorially chosen related terms in the reader's language. Category pages held out of the search index are also kept out of the sitemap.
Your analytics choice controls visitor tracking
The public Crawl Foundry website uses the shared monitoring script. Visitor tracking starts only after you allow analytics, and changing your choice stops further events. URLs are recorded without query parameters or fragments.
Writer protects drafts through connection loss and generation conflicts
Writer saves edits in order and keeps your latest changes when an earlier save finishes. Offline drafts sync when the connection returns. If browser storage is unavailable, the save status tells you to keep the tab open. Confirmed generation runs continue on the server, with progress available after reloading. Completed sections are saved as a recovery draft. A failed run or a document conflict lets you review that draft and choose whether to apply or discard it. Applying it checks the current document version and saves local edits in history first. Content Quality marks its detailed assessment as outdated while you have unsaved changes. New generation and detailed checks wait for the current text to be saved.
Writer widgets stay readable in narrow sidebars
Content Quality keeps dimension labels, score bars and numbers visible when the Writer sidebar has less room. Pending-check messages move below their labels in narrow widgets, and long labels wrap. Widget headings and confirmation buttons also wrap when needed. Findings place the section action below the text in narrow widgets so both remain usable.
Current guides and a clearer documentation navigation
The English introduction, Quick Start, keyword guides, billing and account help now use the same current article catalog as the other languages. The Quick Start shows the keyword list view alongside the steps for adding your first set. The navigation groups SEO tools into expandable sections and includes the Structured Data guide. Research instructions focus on setup, cost, progress and results. Site Audit explains both usage charges and monthly crawl allowances, and clarifies when older crawl details are compacted or deleted. The German introduction includes the published one-minute overview film. Its local preview loads first; YouTube connects when you choose to play it, and you can close the player afterward.
Repeat domain analyses cost less when backlinks barely changed
When you analyse a domain again, we first check its backlink totals. If the number of referring domains and backlinks has barely moved since the lists were last fetched, the anchor texts, referring domains and most-linked pages are carried over instead of bought again. You pay only for what was actually fetched, which saves about a third of the provider cost of a repeat run. The quoted price stays the maximum. The link profile then says which date the lists come from. Lists older than 90 days, a higher plan limit or a clear change in the profile always trigger a fresh fetch. For a one-off analysis you can tick "Fetch backlink lists again" to force it.
Domain Overview lists stay sorted when you load more
Sorting a list in Domain Overview now orders every row the analysis stored, not just the rows already on screen. Load more continues in the order you chose. This applies to keywords, top pages, the keyword gap, competitors, anchor texts, referring domains and most-linked pages, on the overview and in the detail tabs. Loading more referring domains no longer crashes the page. Each list shows how many rows you see out of how many the analysis holds, and how many the data provider knows in total when that is more. The keyword table starts with the essential columns; add others from the Columns menu. Hover a column heading to see what it measures. The Rankings, Opportunities, Pages and Competitors tabs work the same way, and Opportunities now lists every striking-distance keyword instead of the top 20. Pages appear as readable paths without the repeated domain. A small marker flags links to another host, such as www or plain http. Encoded domain names and anchor texts are shown as people write them. Open a referring domain for its authority, its links to your domain and which other analysed domains it also links to. From there you can analyse the domain directly. A badge in the list shows the same overlap. The traffic comparison lists domains as rows, so long names stay readable. Your own domain is always included. The positioning map labels its axes in whole steps such as 100, 1K and 10K. The growth charts start shortly before the first measured activity, fill each row without gaps, and series that stayed at zero are summarised in one line.
Clearer page and competitor cards in Domain Overview
The Pages and Competitors tabs open with two clearer cards. The first ranks your strongest pages, or the competitors with the most shared keywords, with a value bar, the change since the previous analysis and a small trend. The second shows which pages or competitors gained and lost keywords against the provider's previous month, as bars on either side of a zero line. For pages, a line below adds the totals for the whole domain. An entity that was missing from an earlier list now shows as unknown instead of looking like a data error.
One organic traffic number in Domain Overview, now with Position Tracking
Domain Overview now shows one organic traffic number instead of three that disagreed. It is the sum of the traffic of every keyword in the analysis: search volume times the click-through rate of the position, with the same curve in the header, the keyword tables, Pages and Opportunities. Next to it you see how many keywords it covers, and the data provider's own estimate sits in the info tooltip. If Position Tracking measures the domain in the same market, positions measured in the last seven days replace the estimates everywhere, and so does their traffic. A small mark shows which positions were measured, and the optional "Measured position" column also shows older measurements as stale. Tracked domains get a new Tracking tab with the familiar Position Tracking charts: top placements over time, the keyword-by-date grid, the movement distribution, and a list that puts every measured position next to the provider's estimate. From the keyword tables you can send selected keywords straight into the tracking wizard, preset to the analysed market.
Plan your first month with real top-up packs
The homepage calculator uses the available top-up packs and shows how much balance remains after the estimated data actions. Its wider volume range also covers months where a paid plan costs less than top-ups. The estimate starts with a new organization's one-time EUR 1 trial balance. Pricing explains which amounts exclude tax and that only the net top-up amount is credited as balance. German top-up amounts now use consistent currency formatting. The contact form checks the same field rules as the API and identifies invalid input before submission. The skip link follows the page language. Newsletter forms recognize an existing subscription without promising another confirmation email. They offer the German and English editions and identify the email field for screen readers. The changelog shows the day of each release.
Clearer domain charts and readable ranking columns
Page and competitor charts skip saved analyses that have no values for the displayed series, connecting the remaining measurements. Measured zeroes stay visible, and a missing individual page or competitor remains unknown. Long SERP feature lists in the rankings table are shortened to fit their column. Hover over a list to read its full text without overlapping the next column. Domain Overview groups its tabs and analysis controls in one toolbar on wide screens. Open the domain name for market and tracking details. On smaller screens, the controls wrap below the tabs.
Follow your website's Google index coverage
Google Index discovers your website's sitemaps and checks their URLs in a saved website check. The page shows index coverage, a history of completed checks, recent Google crawls and changes to indexing or HTTP status. You can close the tab while a check runs and return to its progress and results. URLs from previous checks remain monitored when they disappear from a sitemap, so later checks can reveal lost index coverage or a changed HTTP status. Connect your Google account to load the Search Console properties you can access. A single matching property is selected automatically. If several properties match, choose from a list without entering a property identifier. Connecting or switching accounts returns to Google Index for the same website. If website ownership still needs verification, the page explains the next step before offering a whole-site check. Admin-approved websites are accepted too. The coverage summary and URL table distinguish indexed URLs, exclusions, URLs that Google does not yet know and unavailable answers. The first check provides a baseline; later checks reveal changes. If Google's quota is reached, the saved check pauses and continues after the quota wait. An explicit minute limit triggers a short pause; daily or unspecified quota errors keep the longer pause. The next retry time stays visible with the saved progress.
The Action Center opens with your priorities
The Action Center now starts with your current situation and recommended next steps. The evaluation date, data-mode label, and separate scoring panel have been removed from the top of the page. Help below the recommendations explains prioritization and provides access to your business context. Optional AI checks still require a separate click at the displayed price. Failed checks show an error, while the existing rule-based recommendations remain available.
Clearer admin settings and more reliable visit tracking
Admin balance adjustments and embedding settings show translated titles and labels with the names and values they need. User profiles show initials when a profile image fails to load. Visit tracking limits repeated attempts after a failed request and preserves recorded interactions when you change pages.
Choose how the public website measures visits
The public website now asks before loading visitor analytics. You can reject analysis and withdraw through the privacy settings on every page. Your choice lasts 180 days and applies to the public website. Ahrefs analytics loads only when enabled for the website and after consent. Analytics excludes URL parameters and fragments. The privacy and cookie policies explain the services and preference cookies. Contact and privacy requests no longer require a consent checkbox. New newsletter emails use unsubscribe links without an email address in the URL, with a clear unsubscribe action. The launch video has a separate consent action and a control to stop YouTube. Blog calls to action have clearer contrast, and customer profile selectors have full, separate touch targets. Plan cards show their monthly billing period. Plans and top-ups display German prices including VAT alongside net amounts and credited balance, with a notice about country-specific tax treatment.
Set up SEO with AI agents in four steps, skill included
A new Learning Center guide shows how to let Claude Code work with your real SEO data. You connect it to Crawl Foundry over MCP, install the free crawl-foundry-seo skill, start with a task that buys and changes nothing, and decide in the agent budget what the agent may spend. The skill is a single, readable SKILL.md file. It tells the agent to use only numbers that a tool returned, to read stored data before buying anything, to state costs before paid steps, and to publish nothing without your approval. You can download it as a file for Claude Code or as a ZIP for the Claude app. Every command on the page is ready to copy, and the connection command is the same one the Agent Platform shows.
Site Audit shows clearer limits and markup coverage
Before a domain is verified, its settings stay within the quick-check limits: up to 100 URLs and depth 2, with rendering disabled and robots.txt respected. Larger settings require verification, including when they are submitted directly to the server. Creating an ownership proof also works when you begin with a larger crawl draft. Verified domains use their current plan limits without automatically increasing saved crawl settings. Changing the homepage to another domain shows fresh ownership instructions. Saved ownership changes take precedence over earlier verification feedback. Larger crawl drafts now receive a matching cost estimate before saving, and start confirmation waits until the estimate has loaded. Structured Data coverage now counts pages that have markup analysis evidence. Failed or unfetched URLs no longer inflate the number of pages reported as having no markup. Older crawls without validation results still show that evidence is missing. Paired performance panels now stretch to the same height, including the empty trend on a first run. The domain menu also tells you when you are already on Sites.
Reliable crawl results and a cleaner Internal Links explorer
Site Audit now handles redirects to trailing-slash URLs without repeating the same request. Crawls can finish collecting page titles and link evidence for those pages. Site Audit processes larger link findings in small batches when repeated navigation fills the initial sample. Incomplete evidence stays clearly marked. The Internal Links explorer keeps URLs from different domains distinct, so selecting your homepage loads its own incoming, outgoing, and external links. Page navigation sits inside the compact page list and appears when more pages are available. Partial crawls and failed page fetches now show clear explanations. Pages without a collected title use their URL as a label, and missing outgoing evidence stays unknown instead of appearing as a confirmed zero. Incoming links collected from other pages remain available.
Product activity overview for our operations team
Our operations team can inspect visits, measured active time, and interactions in internal user records. The view covers product areas, offers 7, 30, or 90-day periods, and can filter the recent timeline by area. Time measurement pauses when the tab loses focus or becomes hidden, and after one minute without interaction. Interaction counts distinguish clicks, form actions, changes, and tab switches without storing entered text. Historical time is not inferred. MCP and API activity distinguishes admitted calls from server audit actions. A permitted request does not imply a completed operation, and browser dwell time is unavailable for those transports. Bounded reads identify partial totals. Existing AI usage and cost records remain available beneath the product activity.
More reliable AI keyword searches
When the shared AI connection is configured, OpenAI keyword searches use it alongside other AI features. Research continues to show actual search queries, and charges use reported costs including cache discounts. Already paid usage remains recorded when a search fails. Gemini searches keep their existing connection because the shared API does not currently expose their search queries.
Pre-Questions included in keyword results
AI fanout results now include prepared questions by default, labelled as Pre-Questions. You can select and import them alongside observed search queries. When a prepared question also appears as an observed search query, one keyword shows both sources. The source filter remains available to narrow the results.
More questions to start keyword research
Question preparation can now generate up to 40 Pre-Questions in one AI call. The questions cover more subtopics, decisions and practical problems while avoiding repeated formulations. You choose which questions continue into AI fanout. The existing preparation result limit and AI token budget remain unchanged.
Find Crawl Foundry on YouTube, LinkedIn and Instagram
The website footer links directly to the Crawl Foundry profiles on YouTube, LinkedIn and Instagram. The About page groups these channels together and links to the founder’s personal LinkedIn profile. German readers can also play the existing product video on the About page. Its preview comes from our website; the YouTube player loads only after you press play.
Clear navigation on the Agent, MCP & API page
Opening Agent, MCP & API now highlights only its sidebar entry. Workspace links stay listed once while the workspace loads and when you switch pages. The full connection label remains readable, with Organization shown underneath.
Stable hover in domain history charts
Domain Overview history charts no longer throw a browser error when the pointer moves over the total line in the ranking distribution. Individual ranking bands still highlight on hover, while the total line remains visible as context.
Expired-domain inventory opens reliably
The expired-domain database loads listings again when the read service uses the previous filter contract. Existing search, sorting and pagination stay available. An enabled Common Crawl filter is never silently removed.
Release reserved keyword research budget after rejected searches
Prepared for the next release; availability depends on rollout. When a search request is explicitly rejected before execution, keyword research now records that no provider cost was incurred. Failed runs can then release their reserved EUR Balance. Requests with an uncertain outcome still require reconciliation before their reservation can be released.
The newsletter reaches every subscriber in one language
Subscribers who signed up in German get the German edition. Everyone else gets the English edition, including readers who signed up in French or Spanish or without choosing a language. Before, those readers could miss an issue. Each subscriber receives exactly one edition, never both.
Manage Agent, MCP and API access across your organization
Agent, MCP & API now has an organization page, clearly marked in the sidebar. Workspace access appears in one list with its workspace name, and changing workspaces keeps the page and selected tab open. Creating an API key requires an explicit workspace selection. Permissions and budgets remain tied to their workspace, and removing workspace access does not disconnect the entire OAuth connection.
Reliable redirects in Site Audit
Site Audit follows redirects to pages with a trailing slash correctly. Page and resource checks preserve the target URL and its query parameters, preventing repeated requests and failed checks on affected sites.
See the free SEO offer at a glance
The English and German homepage headings now say “SEO for free” and “SEO kostenlos”. The next line still explains that fresh data costs extra, so you can see the free offer and its pricing boundary together.
Backup providers for page crawling
Page crawling for Brain, content workflows and Site Audit now uses the same ordered backup providers. When a page fetch fails, the next configured provider can take over within a bounded request time. Site Audit retains actual HTTP errors as findings and requires full HTML for technical checks. The Admin crawler test shows the provider attempts and opens with Crawl Foundry already entered.
Clearer AI usage overview for our operations team
Our operations team can review calls, tokens, provider costs, and daily trends in the internal AI usage view. It uses the same usage records for the selected 7, 30, or 90 days, with recent calls and comparisons by recorded model and feature.
Agent Platform and MCP requests load again
The Agent Platform page and MCP endpoint now load without the server startup error that blocked requests. Existing access and authorization checks still apply.
Deep checks during a rubric service outage
Deep content checks can continue through a configured replacement model when the usual rubric service is unavailable. Those results are marked as partial, so their scores are not mistaken for fully comparable assessments. In the Action Center, a retired decision endpoint now hands the same request to the existing model. The confirmed price ceiling for a deep check stays the same.
Registration emails for our operations team
Our operations team can receive an email when a new user account is created. The notification goes to the platform's configured recipients and includes the account name, email address, registration time, and a link to the internal user record.
Admin navigation and Writer sources simplified
The Admin navigation no longer includes the old Research Sources page. In the Content Writer, the sources panel now focuses on the project sources you add for the article. The Brief continues to show its current evidence and keyword data. Old Research page links lead back to the project overview.
Clearer partial totals in billing history
Billing history now marks totals as partial when older transactions or payment amounts are missing. The net paid figure no longer suggests a minimum when refunds could reduce it. This makes it easier to distinguish the amount shown from a complete account history.
See when monthly usage is being recalculated
When older usage records need repair, the usage page now holds back monthly spending, feature totals, and charts until the figures are rebuilt. Your EUR Balance remains visible. The monthly figures return when the repair finishes. Paid work continues to be charged to the EUR Balance during the repair. If historical activity needs manual review, the monthly figures stay unavailable instead of showing an outdated total.
More reliable EUR Balance and agent spending limits
Top-ups and refunds now keep your EUR Balance aligned with the value credited for a payment. A partial refund removes only its share of that credited value, including when the payment contained tax. If a refund leaves a negative balance, the next top-up reduces that amount euro for euro. Work already underway cannot erase that negative balance when its cost is settled. New organizations show their starter credit in the balance history. Late or repeated payment events no longer apply the same refund twice or replace a newer subscription status. When a subscription recovers, its matching payment reminder closes. A failed payment in one organization no longer suspends its owner across other organizations. For agent connections with automatic spending, each paid content tool call has one budget reservation. If the selected model could exceed that reservation, the call stops before reaching the provider. Repeating its completion cannot release the reservation again, and another connection cannot settle it. No settings change is needed. Bulk keyword imports now apply one charge for the keywords they successfully enrich, within the amount reserved at the start. The charge uses the exchange rate and markup captured with that reservation. Provider costs for missing or unsaved results are tracked separately and are not charged to you. If a charge cannot be recorded immediately, it stays queued for recovery. Configured daily organization spending limits now include provider work paid by the platform. Repeated cost records must match the original amount and billing mode, so a changed retry cannot silently skip a charge or alter the cost history. Admin Billing shows platform costs without adding customer revenue. Recorded costs from failed provider requests appear too.
Find auction domains in Common Crawl
The Expired Domain Database can filter current auction listings to domains found in a completed Common Crawl domain graph comparison. A domain can appear there even when the graph records no incoming links to it. The separate referring-domain filter lets you narrow the result further. The comparison uses historical graph data. It does not check whether linking pages are still live. When a current auction snapshot has no completed comparison, the database says so instead of presenting the missing data as a result.
One overview of every feature and where it fits
The Features page is rebuilt. An example workspace at the top switches between keyword data, site evidence, priorities and AI agents, so you can see the product before you open a single feature page. Below it, all eleven features sit on the five steps of the SEO loop: find, keep, inspect, decide and watch, with AI agents able to run every step through the MCP server. Every feature has its own card with a link to the detailed page, and three questions point you to the best starting point: Keyword Database, Site Audit or Action Center.
Add autocomplete suggestions to your Keyword Database
Autocomplete suggestions in a keyword detail can now be selected and added to the Keyword Database together. The add dialog lets you review the selection and choose whether to request keyword data afterward. The data request still shows its price before you confirm it. If you open a keyword that has no saved entry, the page explains why and offers to add it directly.
See your SEO data as interactive views right in the AI chat
When your agent reads Crawl Foundry data, assistants that support MCP Apps now show it as an interactive view in the conversation instead of a wall of numbers. Nine views cover the workspace overview, Site Audit, rank tracking, keyword research, the opportunity queue, Search Console performance, AI visibility, domain and backlink analysis, and competitor gaps. Each one starts with a clear verdict: a score from 0 to 100, what it means and the next step. Click a tracked keyword to see its position history, a keyword to see its search volume over time, an issue group to see the affected URLs, or an AI visibility run to read the answers and the sources they cite. These details come from your stored data and cost nothing. When data is missing, the view says so and offers the next step, such as starting a Site Audit or connecting Search Console. A value that is not stored is shown as unknown, never as zero. Paid actions show the estimate, the maximum charge and the approval status before anything runs, and site verification shows the exact DNS, meta tag and file values to add. The views follow your assistant's light or dark theme and appear in German or English, matching its language.
A smoother start with a new organization or workspace
Setting up a workspace now stays on its final steps after the new workspace appears. You can start a website crawl, skip that step, and then continue to the dashboard. A full website address also becomes a valid starting point for the crawl. Invited members now see when an organization admin needs to assign them a workspace. Accepting an invitation takes them into the invited organization, including when they already belong to another team. When a person belongs to more than one organization, setup uses the organization they selected. Different organizations can use the same workspace address. If an invitation was sent to another email address, the account switch signs out the current account before asking you to sign in with the invited address. Workspace names written in other alphabets can also be used during setup.
Compare Crawl Foundry and OpenSEO by the bill, not just the feature list
The two OpenSEO comparison pages have been rebuilt so you can see the difference instead of reading about it. Both now open with the first month's bill from each tool side by side: OpenSEO's hosted base plan against Crawl Foundry's free plan. A second chart shows a full year. OpenSEO bills every month, busy or quiet. With Crawl Foundry you top up only in the months you pull fresh data. Every figure still comes from the dated facts table with links to the official sources. The pages also show what Crawl Foundry brings with a link to each feature page, a short switching guide with your keyword import, and the most common questions about switching. They scroll as calmly as the homepage and work just as well on a phone.
See what your AI agent can do, and scroll the product pages without stutter
The AI agents page now shows the agent system at work: an example session with each tool call and what it cost, the two ways in (the SEO MCP server for agents and the SEO REST API for your own software), all 25 agent tools with their cost class, and the checks every agent request passes, from the workspace to the budget to publishing, which always needs your approval. The homepage, Keyword Database, Site Audit and Action Center pages also scroll more calmly. Animated elements no longer jump when a section starts moving, and the headline no longer disappears and reappears while the page loads.
Guides that walk you through SEO step by step
The four Learning Center guides are rewritten for people without SEO experience, such as web designers. Each guide explains its topic with an everyday comparison, shows three rules as good and bad examples, and walks you through the clicks in Crawl Foundry with a simple drawing of each screen. Every step ends with a clear "done when" check, and a short glossary explains the terms you need. The new guide "SEO for web designers: start here" explains what SEO is, which four jobs Crawl Foundry helps with, and whether you should begin with an audit or with keyword research.
Smaller Brain document uploads with a clear limit
Brain document uploads and file imports accept files up to 10 MiB by default. An Admin can adjust that limit in the Limits Control Center, up to a technical maximum of 16 MiB. Files above the current limit are rejected before upload. Uploaded files are tied to the user's Workspace before Brain can parse or attach them. Files from another Workspace cannot be reused in an upload.
Find out who Crawl Foundry fits and which path to start on
The Customers page now shows who an on-demand SEO tool is made for: from a web designer with their own site and a local business to shops, founders, freelancers, agencies, in-house teams and people working with AI agents. Filter the groups by how often SEO comes up in your work, and see which plan each group typically starts on. There are still no placeholder logos or invented quotes; real references appear only with the customer's agreement. The Solutions page has been rebuilt as one overview of all six paths. Three questions point you to the right team or job page, each path shows where it starts and what you get, and a matrix shows that every path runs on the same keyword database, fresh data, Site Audit and Action Center.
The Keyword Database page scrolls smoothly
The Keyword Database page no longer stutters between the opening and the first chapter. The moving background now needs about 70 percent less graphics work while you scroll, which matters most on high-refresh screens and laptops. The keyword chips circling the headline stay sharp without a blur effect, and the reading-progress bar no longer slows scrolling. The old roadmap outlook at the end of the page is replaced by a map of what the database connects to today: keyword research, competitor research and Google Search Console on one side, and the Action Center, rank tracking and AI agents that work from the same rows on the other.
Bring an existing article into your agent's content workflow
Your agent can start with finished HTML, Markdown or a URL from your workspace website, creating an article project with a free quality score. It can request a priced deep check, review suggestions and preview or apply selected changes. The free score remains available when the deep-check evaluator is unavailable; paid checks and suggestions show their price before starting. Saved article versions can be listed, compared and restored. Revision checks protect newer edits from old agent requests. A completed Site Audit of the same website can prepare the Writer without another crawl. Ready or running crawls are reused; new crawls require a verified website and the existing Site Audit price and approval flow. The agent can read saved context checks for missing business facts, firsthand examples, voice rules and topic evidence. It can duplicate accessible projects, inspect briefs and optional competitor-page evidence, mark useful headings, and run the free link-quality check against recent crawl observations. Heading decisions remain tied to current brief evidence. For new articles, edits to goals, audience, questions and notes appear in the same brief as app edits. Rewriting headings, expanding sections or refining outlines offers priced previews and review before saving, with outline revision checks. Website URLs and text notes can be added as project sources; their processing status appears in saved project details. Generated SEO titles, descriptions and slugs can be reviewed before saving, and edited values can be saved without regenerating them. For saved drafts, Optimizer v4 provides the same section plan, price confirmation and section-based rewrite as the app. The agent can review bounded changes and scores before saving a new revision.
More reliable content work through MCP
Agents can now send an article directly to the quick content score, including a Markdown draft. Quotation marks and ampersands no longer trigger a punctuation finding on their own. The score also recognizes common plural and hyphenated keyword forms, and gives a partial match to title words separated by other words. Content optimization now checks that the rewritten article is complete before it reports a score or saves a revision. A Markdown draft comes back as Markdown, with headings, links, lists and emphasis retained. The before and after scores state how many dimensions the quick check measured; editorial rubric dimensions still require a deep check. Comparison and checklist articles can declare their article type so appropriate list density is used. An agent can prepare a workspace website through the existing priced Site Audit flow. Once that audit has usable page evidence, the Writer recognizes it automatically for the same website and workspace. Returning to Content Generation then continues past preparation without a second crawl. If the paid deep check is unavailable on the deployment, the MCP tool explains why and leaves the free quick score available. The Writer also disables the paid check when its evaluator is not configured. Metadata generation, link insertion and optimization remain metered operations under the connection's `content:read` scope and the workspace agent budget. A project deep check needs `content:write`. If that scope is missing, an Organization admin must reauthorize the connection with Content access before the agent retries. Invalid inputs now identify the field to correct.
One newsletter a month, written by hand
Newsletter subscribers get one newsletter a month. We write it ourselves and pick the blog articles, glossary terms and product updates from that month that are worth your time, so your inbox gets a short selection rather than a full list of everything we published. No automatic recap goes out on top of it.
See what stays free before you pay for fresh data
The Owned Data page now shows how the pricing model works instead of listing it. Scroll through your keyword inventory, its lists and tags, and the CSV export, all in the free core, and watch fresh data join only when it is requested. An example estimate lets you choose which data actions a run should include and shows the cost and the remaining starter balance before you confirm. A timeline shows that a top-up adds balance without changing your plan, and that purchased balance stays available after you cancel.
Find your way into SEO faster
The Learning Center now shows every learning path once, as a numbered card in the recommended order. Each card starts with a plain question, such as whether you want to know what your audience searches for, and tells you the level and how long the path takes. If you're new to SEO, the button at the top takes you straight to step 1. The Resources page now starts with three clear entry points: learn SEO, look up a term, or check your website for free. The blog, documentation and changelog follow as short links, then the newest articles. Both pages load without scroll animations, so the text stays in place while you read.
Pricing now starts with top-ups, not a subscription
The homepage and the pricing page now show what most teams need: the free plan plus one-time top-ups. Every euro you top up becomes one euro of balance. The paid plans are still there, and both pages say plainly when one pays off: when you use more than about €31 of data a month, or when you need more than three workspaces. The cost calculator on the homepage follows the same rule. It now recommends the cheapest way to cover your month, so a €10 month suggests a top-up instead of a €30 plan.
Optional workspace domains can be removed
Clearing the domain in Workspace settings now saves the empty value. The former own domain is also removed from tracked domains, so it no longer appears as the workspace's current domain.
Find workspace recovery in the account menu
When a workspace needs a selection, adjustment, or recovery, the account menu now shows a highlighted link to the Workspace Adjustment Center. The notice no longer takes up space above every page. Open your profile in the bottom-left corner to review the affected workspaces and available actions.
The Action Center page reads cleanly on laptops and with screen readers
On laptop screens, the Action Center page no longer breaks words in the middle. The headline, the signal cards next to it, and the chapter headings beside the demos now keep every word whole. Between 1024 and 1280 pixels, the opening stacks the headline above the demo instead of squeezing both side by side. In the goal demo, the selected goal is readable again, and screen readers announce the new order of the queue when you switch goals. The optional data sources now link to the pages that explain them: rank tracking, Google Search Console, and competitor research. Small grey labels across the page have more contrast, and the Site Audit and SEO Monitoring links at the end are easier to tap.
Your AI agent can now score, link and polish articles
Agents connected through MCP or the API now reach the same content tools as the Writer. An agent can score a text for free and get the findings with a next step for each one. It can write the SEO title, meta description and slug, place internal links to pages from your own website, and rewrite a text against its weak spots or your instructions. The tools work on a stored project or on text the agent brings along, as HTML or Markdown. Links come back in the same format, and the wording stays exactly as it was: a result that changes the text is thrown away. For a project, the agent can save the result as a new document revision instead of overwriting anything. Each paid call runs one model request within your workspace agent budget and reports what it cost. Creating a project for an agent is also simpler now: settings the Writer no longer uses are ignored, and the outline step does the research that used to need a separate brief.
New pages for AI agents and the Crawl Foundry MCP server
Two new pages explain how to let your own AI agent work in Crawl Foundry. The AI agents page shows the five kinds of work an agent can take on, from research and planning to content and verification, and the limits it works within: one workspace per session, permissions you can revoke, an agent budget with hard ceilings, and publishing only after your explicit approval. The MCP server page shows the connection address, the three setup steps and the tools with guided setup: Claude, Codex, Cursor and VS Code. It also covers sign-in without an API key, costs from your existing EUR balance, and the REST API for your own integrations. Both pages are in the Platform menu and the footer.
Review global crawler limits in one place
Admins can now review and change the global numeric defaults for Crawler and Brain in the Limits Control Center. The Crawler settings page shows those values and links to the central editor; its robots.txt and re-ranking switches and health checks remain there. Workspace settings can still override the global defaults. Admins can lower six safety ceilings for crawl size, parallel work, and Brain search results; these apply to both sets of settings, including older saved values when a job runs. The editor rejects values outside the allowed ranges and similarity thresholds in the wrong order. It also asks for a reload if another Admin saved changes first.
Find your way around the homepage faster
On wide screens, a slim chapter rail at the left edge now shows where you are on the homepage. Hover over it or reach it with the keyboard to see every chapter and jump straight to Site Audit, pricing, or the questions. The free tools on the homepage now open their tool pages directly, and the pricing overview ends with a way to start for free. Two new answers explain what Crawl Foundry costs and how to check a website without an account.
The Keyword Database page reads better on every screen
On large screens, the chapter text beside each Keyword Database scene is no longer dimmed by the background shading, so every step reads at full contrast. On laptop-sized screens between 1024 and 1280 pixels, the sample tables and the keyword detail panel switch to a compact layout, so statuses, numbers, and search intent no longer break mid-word. On phones, the floating keywords behind the headline stay hidden when motion is reduced, and the newsletter form stacks the email field above the button. The Spanish and French newsletter texts now use their correct accents.
Crawl Foundry compared with OpenSEO
If you are weighing Crawl Foundry against OpenSEO, two new pages give you a fair answer. The OpenSEO alternative page is for readers who want to switch; the head-to-head page is for readers comparing both by name. Both start with a clear recommendation for when each tool fits, followed by a dated table of facts with links to the official sources: getting started, whether a subscription is required, data costs, client workspaces, hosting, Google Search Console and AI agents. Three steps show how to bring your keyword lists along. You find both pages in the Solutions menu under Compare, in the footer and on the pricing page.
Plan changes stay on one subscription, and workspace selection no longer blocks you
When you already have a subscription, choosing another plan on the billing page now changes that subscription. Polar charges or settles the prorated difference, and you keep one subscription instead of two. If a payment is overdue or your plan is about to end, the page tells you and opens the customer portal to sort that out first. After a downgrade or cancellation, you still have 30 days to choose which workspaces stay active. If nobody chooses, the most recently used workspaces stay active and the others become read-only. Nothing is deleted, and you can change the selection at any time.
The platform overview is easier to read and to navigate
On wide screens, the platform overview now has a slim chapter rail at the left edge. It shows where you are and jumps straight to Keyword Database, Site Audit, Action Center, SEO Monitoring, Workspaces, or the feature map. Every chapter can also be linked directly, for example /platform#site-audit. The Action Center demo uses larger, higher-contrast text, and screen readers now announce the new order when you pick a goal. Long headings no longer break in the middle of a word, workspace cards stack on phones instead of hiding behind a sideways scroll, and descriptions across the page are easier to read. The closing call to action has a cleaner animated frame.
The homepage shows what SEO data costs before you sign up
The homepage now opens with a website field. Enter your domain and the free AI Crawler Checker opens with the address already filled in, so you can run a first check without an account. Next to it, the four product layers from Keyword Database to Action Center play through as one stack, with a receipt that prices each sample data action. A new calculator below the hero lets you set keywords, SERP snapshots, and workspaces for a typical month, then shows the estimated data costs and the smallest plan that covers them.
A new Google Search Console page and a clearer end to the homepage
The Google Search Console integration page is rebuilt from scratch. Type your domain to see which Search Console properties cover it, read exactly what the read-only connection may and may not do, and see the four connection states you will find in the product. You can start free with that domain straight from the page. On the homepage, the section on how Crawl Foundry's evidence works now pairs three principles with a sample receipt that shows scope, price, data state, and what the data does not prove. The closing section shows a receipt for getting started that adds up to €0.00 and lists what you only pay once you start it: a single data action or a one-time top-up. A free website check sits right beside it. The footer now ends with a way to start free or check a website first.
Coding agents and social networks can open your signed links again
The Site Audit agent endpoint works for the tools it was built for. When you copy the request for a coding agent, the agent can now fetch the report with its token. Before this fix, the app asked the agent to sign in first, so the request failed. Images in scheduled social posts reach the network again. The network downloads each image through a short-lived signed link, and that link no longer asks for a login. Both links still check their own token, and they expire as before.
The Site Audit page works better on phones and laptops
On phones, the Start a Site Audit bar at the bottom of the Site Audit page is compact again. It used to grow into a tall panel that covered most of the screen, with its note running down one letter at a time. On laptop and desktop screens, the three investigation lenses (Page Explorer, Link Explorer, and Site Tree) now stay fully in view while you scroll through them, so the Site Tree is no longer cut off at the bottom. Their descriptions are easier to read, and screen readers now reach all three lenses. The four steps of the audit loop sit evenly inside their frame, and the hero text stays clear of the crawl graph behind it.
Plan upgrades, data deletion and the privacy policy are clearer
When you upgrade your plan during a billing period, the extra balance now arrives with the paid upgrade invoice. You receive the difference up to the balance of the highest plan you paid for in that period. Switching plans back and forth no longer changes it. Deleting a workspace or an organization now also deletes its research reports and research results. Before, some of these records could stay behind after the workspace was gone. The privacy policy now identifies our hosting and rate-limit services and explains their roles. The database and background processing run in Germany. Business customers can request a data processing agreement (DPA) under Art. 28 GDPR by email.
Workspace settings now focus on active controls
Workspace settings now show the default market for research and position tracking. Content language and target score remain project choices. The workspace form no longer offers a default author style or a text model switch. The external-link blocklist is available directly in Workspace settings. Research seeds and keyword filters are accessible from Keyword Research for the selected Workspace. Older global Crawler and Research settings links return to the settings overview so they cannot edit a different Workspace by accident.
Writer preparation shows crawl start problems
If the crawl cannot start during the first Writer preparation, the page now shows the problem instead of silently staying at zero pages. You can try starting the same run again. An unusually long wait is also visible after reloading the page.
Connected agents can work right away
An agent you connect through MCP or the API can now run paid SEO research and content generation from the first minute. You no longer have to switch on an agent budget before the first run. Until an admin saves their own settings, every organization runs on a default budget: up to 5 € per action, 10 € per workspace per month and 50 € per organization per month. The caps apply to every run, and each run is charged against your balance as before. Publishing is not covered by the budget. In the agent budget settings you can lower the limits, restrict them to research or content, or turn automatic spending off entirely. Connections whose policy you set by hand keep that setting.
Compare source outlines and guide your own structure
The Content Generation brief now compares the readable search pages with your own outline. It shows their heading levels and relative text length without treating either measure as a quality score. Open a page to inspect its saved headings; pages that could not be read remain clearly identified. Mark individual topics as important or ones to leave out. Your choices are saved with the current research and stay optional. When no topics are marked, the readable results provide broad context. The AI uses these observations to suggest an original outline suited to your keyword and page goal, rather than copying competitor headings. The suggestion shows its maximum price before the request. You can compare it with your current outline and choose whether to apply it. Your existing outline is preserved until you accept the proposal.
Explore domains without losing your place
Prepared for the next release; availability depends on rollout. The domain database now stays inside the main app navigation. Open a domain to review its current auction, source metrics and saved research status, then return to the same search, page and column view. Filters and columns open in focused panels, and new domain lists open in a dialog. Shift-click selects a range of visible domains for saving. Each domain profile links to its earlier website versions in the Internet Archive for manual review. The database is now the single view for expired-domain research. Its domain profiles group auction details, provider estimates and saved research status. Separate SEO metric columns show "Not available" until data has been added; these values are distinct from the imported auction metrics.
Check a domain’s link history before you decide
Prepared for the next release; availability depends on rollout. The expired-domain database can add current link metrics to selected domains. In a domain profile, you can also request a monthly history of referring domains and a sample of the domains linking to it. This helps you check whether an imported auction estimate matches the domain’s link history. Each lookup shows a maximum price before you confirm it. Results are labeled by source and date, and missing provider data stays visibly unavailable. A failed lookup does not trigger an automatic repeat purchase.
Clearer limits for content and document imports
Content Generation rejects oversized article documents and Lite import outlines before saving. Brain parsing stops oversized requests, downloads and parsed responses. Existing workflows remain available within their documented limits. Writer source files are attached to an editable Project together with their source record after access, type and size checks. The default is 10 MiB; the maximum configurable value is 16 MiB. Older upload links or storage IDs cannot attach files. Source fields must match their type, and source links require valid web addresses. Active-source quotas ignore deleted records; comments load in pages, and deleted comments do not block replies. Site Audit alerts reject oversized recipient lists before processing them. Content Briefs receives its configured limits when loaded. Media uploads check their 3 MB file limit, with a 4 MB request cap; source URLs must use HTTP or HTTPS without credentials. Context Check rejects blank answers and oversized questions. Older oversized questions require a fresh analysis. Style Profile extraction checks destinations and redirects and limits fetched pages. Source URLs accept up to 2,048 characters; analysis accepts ten samples and 100 KB total, shortening longer articles with a notice. Access and Workspace ownership are checked before analysis. Styles load in pages, searches indicate unloaded results, and saved styles from another Workspace are ignored. The Limits Control Center distinguishes browser display settings from server admission budgets. Out-of-range saved settings do not take effect, and retention increases stay capped until verified safe. Keyword and competitor reports keep visible rows when detail limits are reached, identify partial results, and preserve separate read, query, import and maintenance budgets.
Clearer Optimizer results for unchanged rewrites
If the Optimizer receives an unchanged passage despite a concrete finding, it now reports that passage as not revised instead of saying no changes were suggested. Its rewrite instructions also point to useful detail already present in the article when a finding calls for more specificity.
Clearer optimizer edits for long drafts
Prepared for the next release; availability depends on rollout. When a recent deep check has SERP comparison data, the Writer and Optimizer now reuse that comparison while checking the current draft. Long drafts no longer lose the relevant structure finding during the handoff. If a finding describes an entire section as vague, repetitive or generic, the Optimizer reviews every affected paragraph in that section. Local issues, such as one unsupported claim, still call for a focused edit. It preserves useful details and does not invent sources or personal experience.
Clearer draft recovery and version comparison
The Writer no longer offers an empty editor shell as a recoverable draft. Saved content also clears its local recovery copy when it reaches the document. A failed writing attempt now has a dismissible message with space above the editor. The Optimizer can be folded away while keeping its review state, and comparing article versions opens without an editor error.
Plan the page and review its outline in one brief
The Content Generation brief keeps the page goal, audience, research and outline in one workspace. Edit the target keyword, language and supporting keywords directly. Research starts only after you confirm its displayed price and remains unavailable when it cannot be requested. If a request fails, saved results stay available with a clear status rather than technical provider details. Outline checks remain visible while you edit headings. Word budgets, titles and meta descriptions belong in the Writer rather than a second set of brief fields. Deep quality checks use the saved page goal and audience; editing that context marks an older result as outdated. Observed topics and questions from readable search pages appear ahead of the source list. Each observation links to its page so you can decide what belongs in the outline. Audience and business goals remain optional, and the keyword can be edited beside the title. The source list separates search results from fetched pages. It shows how many selected websites yielded usable text and distinguishes unsuccessful fetches from pages never checked. It does not imply that every unread result blocked the crawler. Older research without content observations shows its status instead of presenting three search-result titles as findings. Pages with a banner state can be read without discarding their article text. Existing results remain available; refreshing a previously unread page requires new research. Older research without captured heading levels offers a refresh action. Sources remain readable during that request, whose maximum price is shown before you start.
Review and add article links together
The Writer link panel combines existing links with internal and external suggestions by default. Separate views remain available. The Writing Assistant card has no nested card or sideways scrolling; choose individual links or apply suggestions together. Clicking an existing link opens its exact position and editor controls. Internal suggestions use observed workspace pages and topic phrases present in the article. They can include own-site pages saved in the Brain or completed Site Audit results; the empty state explains when none are eligible. External research reads candidate pages, excludes the workspace and explicitly marked competitors, and keeps automatically discovered SERP domains eligible until reviewed. Larger source pages are accepted during verification. External searches show their maximum price before confirmation, capped at one SERP request plus one semantic request. Only executed requests are billed. Diagnostic IDs and candidate acceptance or rejection reasons help explain empty results. The combined automatic action checks own-site pages first and adds no more external links than newly added internal links. Without a suitable internal page, it explains why and skips paid external research. You can still run either type separately. When an article already has links, choose to keep them or rebuild editorial text links. Rebuilding preserves contents, contact and navigation links. The original stays unchanged during preparation; review removed and added links before saving. If no suitable new internal link is found, existing links stay. Insertion applies anchor choices to the original article, skips targets already linked when preserving links, and rejects rewrites or changes to existing links. Returning saved content is recognized as the same editor version. A different server edit still opens conflict resolution with the local draft intact.
Deep checks flow into the Optimizer
When a text has no current deep check, the Optimizer offers one before preparing its rewrite. The price is shown for confirmation, and the saved result is reused while the article stays unchanged. Editing the article makes the old check stale. The deep check's embedding work now falls within the same two-cent maximum. Optimizer failures also retain a more specific diagnostic status, while the editor shows one error notice instead of repeating it.
Clearer Optimizer recovery when a run fails
Optimizer errors now fit inside the writing assistant, with readable text and recovery actions on separate rows. When all rewrite calls fail for the same known reason, the error carries that more specific support code so a failed run is easier to diagnose.
A clearer Optimizer from plan to review
The Optimizer now presents its plan, live progress, and proposed changes in a calmer layout with simple section rows. Prices and actions stay easy to find, and long section titles remain readable on narrow screens.
Optimizer starts after confirmation
Confirmed Optimizer runs now use the correct organization for their reserved model calls. The recovery buttons also sit side by side in narrow panels, with labels that can wrap instead of being cut off. The article rewrite action opens a fresh Writer run again. Optimizer results stay available when the panel is collapsed, and the proposed text appears as an inline review without saving it as a draft. Accepting a partial result saves the selected sections against the original document revision.
Opening the Writer no longer starts hidden paid checks
Opening or reloading Content Generation no longer requests AI outline suggestions in the background. Those suggestions were not shown in the editor, so the hidden request has been removed. The Writer also no longer starts a paid deep quality check when a generated draft opens. The check remains available through its visible confirmation step. The free, local text checks continue while you edit.
Article context is easier to inspect in the Writer
The Sources panel in Content Generation now separates project inputs from sources in the linked research report. You can read your saved text, open source URLs, and see which project inputs are still processing or failed. URL, text, file, and image inputs remain available to add before another generation. When adding a URL, you can enter a domain such as `crawlfoundry.com`. The source is saved as a web address without requiring you to type `https://`. If background processing cannot start, the panel now says that the source was saved and offers a retry. Pending sources are picked up again automatically after an interrupted start, and failed sources can be retried from the list.
Concrete rewrite suggestions for your top findings
The Writer's findings can now come with a proposed edit. After a deep check, one click writes suggestions for the most important findings: the passage as it stands, the rewritten passage, and one sentence on what the change fixes. Apply a suggestion and the passage is replaced in the editor at exactly that spot; the rest of your text stays untouched. Suggestions are grounded in your text. A passage is only offered when it exists word for word in the article, and when a finding asks for knowledge the model cannot have, such as your own measurement or a source, the suggestion leaves a marked gap for you to fill instead of inventing a figure. Every round is one model call with a price ceiling you confirm first, and suggestions only ever belong to the text revision they were written for.
The quality score drops the "New contribution" dimension
The Writer's quality score no longer shows a "New contribution" dimension. It measured how far your sentences sat from the pages that rank, and on real pages that distance rewarded drifting off the topic rather than adding substance. We checked it against human grades on 162 ranking pages in English and German and against a control run with unrelated competitors, and it did not hold up. Substance is still judged, through specificity, first-hand experience, honest limits and supported claims, and those dimensions now carry the weight the evidence supports. Saved deep checks stay valid; the retired dimension simply disappears from the card.
Site Audit article and AI search glossary open to search
The article on how Site Audit turns crawl evidence into technical SEO decisions now lives under a Crawl Foundry address in English and German, and old links redirect there. Released articles appear in the new blog sitemap and can show up in search results; the blog overview stays out of search until more articles support it. The glossary also releases six reviewed concepts on AI search: AI Overviews, AI Overview optimization, generative engine optimization, Google AI Mode, cited source optimization and AI hallucination.
Clearer brief page and visible keyword marks in the Writer
The brief page is calmer: smaller numbers, a tighter outline and more room at the sides on wide screens. The structure metrics now speak your language and mark each finding correctly, so a pass no longer wears a warning icon. When the competitor research cannot complete, the brief shows the underlying reason next to the failure code instead of a bare "unavailable". In the Writer, "Mark in text" works again: target keywords are underlined in the article once the toggle is on.
A clearer Content Generation start page and a Writing Assistant in the editor
Content Generation now opens with what the surface does — brief, outline, article, writing assistant — plus direct entry points for a new project or improving your own text, and a table that shows each project's current step at a glance. The article editor reads like a document again: one formatting bar on top, the page at reading width, versions and the AI rewrite in the header. On the right, the new Writing Assistant scores readability, SEO coverage, structure and length locally and for free, lists concrete recommendations you can jump to, and shows how well each target keyword from the brief is covered. The separate external review, approval and share-link workflow has been retired so the writer stays focused on producing the article.
A calibrated quality score for every article
The Writer replaces its local checklist score with one content quality assessment. Readability, structure, scanability, keyword basics and links are measured instantly while you write. A deep check adds calibrated judgements per section: how concrete the claims are, whether first-hand experience and limits are named, how much filler a section carries, whether the opening answers the query early, and which expected subtopics are still missing. The deep check names its price ceiling and asks for a confirmation before it runs, like every paid action. Every dimension shows a 0–100 score with a plain state and a next step. Judgements that are too uncertain are shown as not measurable rather than guessed, and a result only ever belongs to the text revision it was measured on. Findings name the section and take you there. The Content Generation list now carries the shared workspace header with help, problem report and the create action, like Position Tracking and Site Audit.
The Optimizer now works from your quality score
The Optimizer rewrites against the same content quality assessment the Writer shows you. Every dimension below the target of your preset becomes a task for the rewrite, and the findings from the deep check tell the model which section to fix and how: a vague section gets concrete figures, a missing subtopic gets two specific sentences, a contradicted claim is corrected or removed. Nothing is scored twice and nothing is bought behind your back. The deterministic dimensions are measured on the text you optimize, and the deep-check judgements are used while they belong to the current text revision. Presets weigh the sixteen quality dimensions; runs are no longer blocked by a single metric, and custom weights you saved earlier are carried over.
Clearer Writer errors and a lighter Content Generation start
When the Writer cannot start an article, it now says why: a changed document, an outline it cannot read, an insufficient balance, or an expired price each get their own explanation and next step instead of a generic internal error. Unexpected failures keep their support code so we can trace them. The Content Generation page opens with the full introduction only while a workspace has no project yet. Once projects exist, the introduction shrinks to a slim band, and the cards below it explain what the surface does: writing a new article, improving your own text, and finding weak spots while you write.
Prepare and write in one workflow
Start with a keyword, page or text. Confirm a maximum price before writing or optimizing. Duplicate confirmations cannot buy another run; changed inputs need a fresh preview. Choose your website, confirm costs and prepare up to 50 pages with Site Audit. Reuse preparation without extra charges; stop to change domains. Brain and Research lead to the writer, reusing website information and business settings. Drafts survive failed revisions and telemetry outages. Incomplete generation stays labelled. Local checks flag review needs without verifying facts. Draft and history save together; retries reuse the stored text. History keeps 20 recent versions and up to 20 milestones, preserving previously protected versions and small edits. Imported drafts avoid loading loops. Formatting stays left and tools right. A rewrite button opens the optimizer. Page requests have time and size limits. Prepare with search results and up to five competing pages. Confirm the price before starting. Unavailable pages leave other results intact. Labelled excerpts provide topic hints, separate from your own evidence. Agent generation checks research changes after approval. One optimizer replaces Precision Fix; hidden Brain panels stop loading data. Free checks cover 48 links, readability and structure without verifying facts. Unknown targets remain unchecked. Editing clears results. Late responses cannot replace a newer optimization or another project. Cancellation stops further steps; started work can still incur costs. Failed billing is retried without generating text again. Partial optimizations identify unchanged sections. Failed runs preserve your draft.
More reliable scheduled and background SEO jobs
Prepared for the next release; availability depends on rollout. Domain Overview, Position Tracking, keyword research and competitor requests now save their pending request together with the order and its reserved budget. Temporary delivery failures can be retried without creating another paid request, and delayed duplicate deliveries cannot start the same work twice. Overdue Tracking periods produce one current run. Site Audit checks again shortly when crawl capacity is occupied. Paused Tracking orders remain paused, archived results remain available, and requests that cannot start within the retry window end with a visible failure and release their reservation. Domain Overview and competitor results can recover from a temporary saving error without repeating a completed data request. Tracking recovery preserves its full window after a delayed start. Queued Site Audit requests and keyword research recovery also handle competing queued work more reliably. Keyword data and SERP enrichment also retain accepted requests when a connection drops. Repeating the same request reuses its saved job. Already submitted SERP results can be collected after an interruption, and a delayed result cannot charge the same keyword twice or replace a newer job's result. Cancelled enrichment requests reject new result writes.
Reliable detection of updated MCP tools
MCP connections now advertise a schema fingerprint that matches the current tool definitions, so clients can detect when they need to reload their available tools.
Connect your workspace through MCP and API
Agent, MCP & API is now available in workspace navigation. Organization owners and admins can manage connections and API access without a platform administrator account. Other members can view setup instructions and available tasks; managing access and changing budgets still requires the appropriate organization role. MCP setup links opened from the former application address now use the current Crawl Foundry endpoint for OAuth sign-in.
Stable navigation after opening the Writer
Opening a Content Generation project no longer switches the sidebar, tabs, and header buttons of every other page into their narrow-screen layout until you reload.
Find workspace activity faster
The activity page brings recent operations closer to the top, with compact filters, readable metrics and one activity and spend chart. Open additional analyses when you need the yearly calendar, category trends or spending by user. Operation details retain durations, results and billing information. Consecutive standalone charges are grouped and remain individually inspectable. Charges show their billing status instead of implying a successfully completed job. Search and CSV exports identify their loaded data scope. Historical errors are described as recorded statuses, and unavailable calendar data no longer hides the working activity list. Choose 7, 30 or 90 days, or review activity across 6, 12 or 24 calendar months. Metrics, operations and CSV exports follow the selected period. Report an activity problem beside its visible error code. The draft includes the operation reference, recorded status and time window; loading errors also carry a diagnostic reference. Existing error codes survive merged activity entries, and unclassified failures remain explicitly marked.
Compare competitors with more data in view
Competitive Research now opens with a compact domain comparison and visible traffic shares, ranking distributions and organic versus paid traffic. Organic traffic, keyword and paid traffic histories appear as separate charts with a shared time range and domain selection. Link quality, keyword overlap, keyword gaps and successful pages complete the comparison. The overlap shows actual intersections within the stored ranking samples and states that limit. Requested charts keep their place while data arrives. Loading, partial results and retry actions appear inside each chart, with source status available on demand. The large coverage table and repeated promotional notices are gone. Additional analysis uses the shared launchpad. The competitor catalog uses compact rows, search and filters for active competitors, candidates and domain context. Source details and actions open beside each row. Existing comparison links, keyword navigation and price confirmation are preserved. Scrolling over the competitor table now continues up or down the page when the table reaches its own scroll boundary.
A clearer website overview
The dashboard brings domain performance, tracked rankings, audit results and next actions into a compact overview. Switch tracking schedules, reporting periods and chart metrics without losing your selection when returning from a detail view. Ranking charts use recorded measurements and show missing observations separately from keywords that were not found. Changes appear only when the measurements can be compared. Empty workspaces start with one shared introduction; partially populated workspaces keep their available results visible. Responsive mini-charts now align pointer selection and tooltips with the drawn data. Tooltip labels wrap without colliding with values, and dashboard columns share the row height across populated and empty states.
A clearer domain overview, even with limited data
Domain Overview keeps seven analysis tabs with expandable source details. Available results remain visible when data is limited. Missing measurements stay distinct from zero, gaps stay gaps, and a single observation never becomes a fabricated trend. Sparse backlink distributions use compact summaries. Each tab leads with its evidence: Opportunities shows positions 4-10 and 11-20; Rankings shows Top 3, Top 10 and organic history. Competitors highlights shared keywords and ranking movement; Pages highlights traffic and winning or declining URLs. Comparable saved observations become line charts; otherwise current values remain bars. Monthly SEO history and saved analyses keep separate dates and scopes. Charts appear only with enough measurements and share the available row. Cards align, source details sit in the navigation, and explanations open on demand. Page tables fill their column before scrolling and show traffic changes only when the preceding analysis measured the same URL. Competitors loads more rows in the overview. Opportunities without evaluated matches no longer show unrelated history cards. Saved-history subscriptions remain stable when tabs or results change. The analysis form combines one-off analysis with recurring drafts, price and balance. Paid one-off analyses require explicit price confirmation. Drafts offer weekly, twice-weekly, monthly and custom previews; monthly is the default, with four planned dates. Cadence choices expose weekday or month-day settings where needed. Dates and drafts are clearly inactive: the preview neither persists schedules nor starts automatic analyses. The Orders page shows status counts, search and inline draft management. Drafts can be edited or removed while the page stays open; empty Orders offers a direct action to create an analysis. Calculation details remain expandable.
Schedule recurring domain analyses
Domain analysis orders are now saved with their schedule, approved price per run and run history. Choose weekly, fortnightly, monthly or a custom interval, then pause, resume or archive the order. Archived results remain accessible. Overdue orders are retried after interruptions. Multiple missed dates produce one current analysis rather than a burst of paid historical runs. A higher price or changed keyword allowance requires a new confirmation. The keyword selector shows the scope available to the workspace.
More reliable geographic checks
Starting or repeating a geographic check now starts every measurement, including the initial keyword-data lookup. Retrying the same request reuses its existing jobs. A check without usable results no longer displays a successful completion icon.
Compare geographic rankings in one map workspace
Geographic tracking brings average position, Top-3 and Top-10 coverage into one summary above the map. Select your domain or a discovered competitor to inspect its recorded positions, then choose a measurement point to open the relevant keyword results. Compact filters, a closer map view and a point navigator keep the analysis together. Pending, failed and unavailable observations remain distinct from measured rankings.
Simpler setup for agents, MCP and API access
Choose your tool to see its installation link, command or API example. Access management now brings API clients and agent authorizations together, with permissions available in expandable details. Copy actions show confirmation, saved clients can be loaded in pages, and failed requests remain distinct from empty lists. The task catalog and budget settings use a quieter layout that adapts to smaller screens. Configured access and recorded credential use are shown without implying a successful connection test. Tool choices are now visible buttons, tasks open directly into readable cards, and each saved integration has its own card. Create API access from the toolbar. Reload controls explain that they fetch saved status, and budget controls advance in whole euros. Budget settings now show only the decisions needed for the selected mode, with visible workspace choices and organization defaults in one expandable section. Organization-wide activation is explicit, saved usage stays separate from draft settings, and unsaved changes are clearly marked.
Clearer keyword data and comparisons
Keyword details show the data date at the top right and present competition metrics in a compact row. Bid ranges are displayed as amounts, with a note when the currency is unavailable. Data and sources now opens into a flat overview of market, language and measurement dates. Choose a saved data snapshot from a compact selector to compare it with the current values, without changing the charts above.
A clearer keyword collection header
The Keyword Database now keeps collection size, search volume and average difficulty in one compact summary. The page title identifies the selected collection or list, while metric tooltips explain how filters affect the results. Missing metrics remain distinct from measured values; unsupported summary scores and the currency-less average click price no longer compete for attention. Search has more room, and view controls and tools wrap in clear groups on smaller screens. Add keywords remains the primary action, with quieter access to export, help and support.
Return to your keyword list without reloading it
Opening a keyword now keeps the database list loaded. Returning restores your filters, selected view and scroll position, including keywords further down a large collection. Saved keyword metrics appear without waiting for the separate search-results metadata. Direct links to keyword details still work. Interrupted data enrichment jobs now record a failed outcome. A recovery check also closes old unfinished runs once their reservation is no longer active and no settlement is pending. Saved results and billing records remain intact; recovery does not purchase another data request.
Switch between saved SERP views
Keyword details now offer a list view and a Google-style view of the same saved search results. Switch views without losing your current results page. Browse ten results per page, up to the number actually stored for the selected measurement. Your domain and active tracked competitors are marked in the list. Results beyond the earlier measurement's recorded depth no longer receive an unsupported “New” label. The displayed measurement and its comparison are labeled separately. The latest saved list with more than ten results opens directly when available, with its actual measurement date. Page changes no longer automatically scroll to the section heading. Google-style pagination stays inside the browser frame. Both views highlight tracked domains and load fallback favicons. Available dates, ratings, authors and sitelinks also appear in the list.
Clearer Site Audit reports
Domain and crawl selectors now sit in the Site Audit header, while every report stays directly accessible in the tab navigation. Overview and Internal Links cards align to the same height. Health and category values appear beside simple bars, and click-depth statistics use labeled rows. URL details have a new layout for page context, key metrics, findings and expandable checks. Settings use a compact two-column form, automation activity uses aligned rows, and the crawl request log scrolls within a bounded area with fixed column labels. Returning from URL details preserves the Pages search. Structured Data distinguishes detected markup from validated results. Unvalidated markup no longer receives a clean status, and missing validation totals appear as unavailable. Performance checks waiting to start are labeled as queued. Issue reports now show affected URLs across crawls in an interactive area chart. Statistics shows crawl trends and compact breakdowns directly, including sitemap coverage and internal link structure. Incoming and outgoing links use compact aligned rows, and the URL Schema tab groups types, validation counts and expandable evidence. URL findings load in severity order across cursor pages, so loading more keeps earlier rows in place. Schema findings have their own filtered query, and expandable page checks highlight their full width. Internal Links now scrolls through the complete overview. Its Explorer uses a searchable page list, four key metrics and one placement filter. Automation activity uses compact expandable event rows. Older comparison crawls remain available inside the crawl selectors, without a separate toolbar button. Statistics adds trends for pages outside the sitemap and places weak-link history beside its current breakdown. New crawls also record the average measured HTTP response time of successful internal HTML pages in milliseconds. Older crawls without timing totals remain unavailable instead of appearing as zero.
Start crawls with older saved settings
Site Audit resets saved crawl depths that exceed the current limit to the allowed default, so an older setting no longer blocks the next crawl. Valid depth settings stay unchanged.
Quieter loading in Position Tracking
The overview no longer flashes a partial-results notice while keyword data loads.
Longer reporting periods in Position Tracking
Choose 7, 30 or 90 days, or 6, 12 or 24 calendar months from the shared reporting toolbar. The selected period updates the overview, keyword and competitor queries and remains selected when switching tabs. The overview now starts directly with its metric charts, without an extra performance heading.
Compact tracking orders and saved history
Position Tracking now presents orders in one compact, searchable list with status, last run, next check and estimated monthly costs. Active, paused and archived orders are easy to filter. Actions and saved run history open in a side panel, keeping the overview focused. New workspaces use the shared feature introduction with a direct setup action. Loading and failed requests remain distinct from an empty workspace, and archived results stay accessible.
Stay in the same section when switching workspaces
Switching workspaces keeps you in Position Tracking, Site Audit or your current section instead of sending you to the Keyword Database. When you switch from a detail page, the destination opens the section overview so you can choose a record from the new workspace.
Check the topical fit of an internal link
Action Center can now assess the specific internal link attached to a selected measure. Start the check from the measure details after reviewing the price. The saved result identifies both pages and shows good, conditional or unclear fit separately from the existing priority score. The check uses stored page titles and anchor text; full page content and placement still need review. The optional fast evaluation falls back to the existing AI evaluation when unavailable or uncertain. The original recommendation remains available if evaluation fails. No links are added to the website automatically.
See the correct confirmation after adding keywords
Adding keywords now displays the number successfully added or restored without a translation error. Existing keywords that were skipped are not included in that count. When you add keywords while viewing a manual list, they are also added to that list, including keywords already present elsewhere in the workspace. The confirmation appears after the list assignment succeeds.
Browse large keyword lists without a loading error
Manual keyword lists with more than 1,000 members now load in bounded pages instead of failing before the first results appear. Sorting and combined filters continue to apply as you browse the list.
Blog and glossary in the navigation
Blog, SEO glossary and free tools are now available in the Resources menu on desktop and mobile, and in the footer. Owned Data is also linked from the product menu and footer.
Keep paid tracking results after ending a schedule
Archiving a tracking schedule stops future checks while keeping its saved results and run history accessible. Archived schedules appear in the tracker selector and in a dedicated archive under Orders & Settings. Their reporting window starts from the last recorded run, so older results remain visible. The keyword heatmap and table now use the full page width, one below the other. A single compact toolbar combines search and ranking filters; both views share the same filtered, sorted page of 20 keywords.
Connected performance charts in Position Tracking
Estimated clicks, average position and visibility now connect recorded values even when the measured keyword selection changes. Each tooltip shows how many keywords contribute to that value. Missing values remain gaps rather than becoming zeroes.
One detail page for every keyword
Keywords now open on a shared page from the Keyword Database and Position Tracking. The return button restores the originating view, including its filters and selection. Search volume, difficulty, intent and CPC sit above the demand and click-potential charts. Missing data stays empty. Related keywords, enrichment, tags and saved data versions remain available. Saved search results appear once. Where comparable measurements exist, position changes appear beside each result. Device, location and language remain part of the selected search context. Google autocomplete now treats a successful lookup without suggestions as an empty result instead of a service error. SERP sitelinks sharing the same URL remain visible without duplicate-key warnings.
Optional business-fit evaluation with automatic fallback
Action Center can use an optional evaluation path to suggest how unresolved keywords relate to a customer's offerings. If that path is unavailable or its answer is uncertain, the existing evaluation continues automatically. Confirmed customer choices remain protected. The optional path is disabled by default. The assessment explanation now separates rule-based prioritization from optional AI keyword-to-offering checks. Those checks show when they are running, when suggestions are ready, and when no supported assignment was found. AI results are not presented as ranking predictions or an SEO success percentage. A direct AI assignment test button opens the prerequisites, quoted price and start control together. Opening the panel does not start a paid evaluation. Saving business context now confirms that the changes were saved separately from an AI check. When no keyword assignments need review, the panel explains that internal links and action ordering are not included in this check.
Clearer information about website traffic measurement
Our privacy and cookie notices now explain the use of Ahrefs Web Analytics on crawlfoundry.com, including cookie-free visit measurement and the use of aggregated data. Tracking is limited to the public production website.
Keep Content Studio connected to its publishing workspace
The admin Content Studio, calendar and media library now use the configured publishing workspace independently of the workspace selected in the app. Adding a channel explains that generation uses the publishing organization’s EUR Balance; reopening a saved edition does not start another generation. Article-based social editions include their source link on supported channels. Missing links block scheduling with a clear message. Scheduled articles can supply their future URL, while delivery waits for the public article. X and Mastodon generation instructions now call for different editorial angles instead of shorter and longer versions of the same summary.
Find technical SEO and keyword concepts from the glossary
The English and German glossary overviews now link directly to 16 reviewed concepts covering crawling, indexing, internal links and keyword research. These articles can appear in search and are included in the glossary sitemap. Filtered views remain available for browsing without creating additional search results.
From starting keywords to reviewed AI search questions
Start with a topic, prepare questions from autocomplete and AI suggestions, then choose and edit up to 100 questions before running AI web searches. Select or deselect recommendations together while keeping your own entries. Each step shows its price; larger selections use the same research endpoint. Direct research also offers AI keyword suggestions without web searches or invented SEO metrics. AI fanout starts with one model and a compact scope. Expanded scope and a second model are optional. The maximum price updates with your selection; native fanout uses a 1.2 cost multiplier by default. Final charges use recorded usage within that maximum. Suggestions and observed search queries keep their source labels and original context when combined. AI suggestions, prepared questions and fanout results include Essentials lookup for volume, difficulty, intent, CPC and monthly trends. The price preview itemizes this lookup. Metrics loaded during preparation carry over into fanout; unavailable values stay empty. Traffic potential remains an estimate. MCP offers the same preparation, source selection and scope. Prepared questions stay within their original workspace and market. Reports distinguish starting questions from observed search queries and initially filter to AI searches. Missing coverage and failed metric lookups have separate explanations. Successful lookups with no data are reused. Runs without observable searches no longer report prepared questions as a successful fanout. Ranking history reads database pagination correctly. Results put keywords and metrics first, with compact source rows and explanations on demand. Original questions and AI searches open from the top row; long questions stay out of the heading. The research form retains the full price breakdown.
Add your first keywords from the empty database
The Add keywords button in an empty keyword database now opens the keyword entry dialog, so you can add your first keywords directly from the empty state.
Shorter ranking tables and clearer search-demand analysis
Position Tracking tables now show 20 entries per page. Sorting and filtering apply to the loaded analysis before it is split into pages. Discovered competitors can also be searched by domain. The keyword heatmap stays visible alongside the detailed table, and page metrics open directly. The shared period selector offers 7, 30 and 90 days within the available history window; 30 days remains the default. The overview keeps its three timelines and adds search-demand distribution by ranking range, plus keywords on positions 11–20 ordered by known search volume. Unknown search volume stays separate from zero, and incomplete measurement data remains visible. Search volumes are summed keyword estimates, not unique visitors or a forecast of additional clicks. Large, partially loaded analyses identify their loaded keyword basis.
Faster keyword analytics updates after imports
Production keyword analytics now process queued updates in larger batches and check for new work sooner. Imports no longer inherit the low-throughput settings intended for the development environment. Ranking and opportunity reads also check whether their source updates have reached the analytics store. Failed updates remain visible until a successful repair, and background processing now has explicit limits on batch size, response size and request duration.
Keep Content Studio schedules intact during status checks
In the admin Content Studio, checking a delivery status preserves its planned send. Retrying an unconfirmed request retains its identity and revision safeguards. Blocking editorial issues prevent scheduling and are checked again before sending. DEV.to editions retain the source article link. Media checks cover missing dimensions, file sizes, video duration and supported channel combinations. The calendar flags later delivery conflicts even when a publication receipt already exists. The edition view explains that service status checks cannot reliably detect edits or deletions made directly on a social platform.
Clearer ranking insights and side-by-side page comparisons
Position Tracking now connects ranking observations with their meaning and a concrete next check. Keyword details let you compare your landing page with a saved competing result, including positions, titles and descriptions. Differences between saved result lists remain separate from claims about ranking causes. Related charts and summaries now share a bounded, responsive layout. Short histories use compact charts with visible measurement dates; sparse keyword charts label their points directly. Heatmaps follow the keyword table's search, sorting and pagination. Landing-page rows include their ranking distribution, and competitor discovery uses compact comparison bars above its paginated table. Topics, directories and keywords without a found URL remain visible together, while competitor groups show where you lead, trail or have no result within the measured depth. Neighboring panels now finish at the same height, including paired charts, keyword tables and stacked topic summaries. Charts and lists adapt to the available space, and chart tooltips can extend beyond a panel without being cut off. Keyword details fill the row whether two or three insight cards are available. All six report tabs remain visible on narrow screens, and keyword search updates without a server navigation for every keystroke. Traffic estimates, average positions and AI citations now show their data limits more clearly. Geographic checks explain the selected point alongside the map, keeping missing measurements separate from poor rankings. Tracking reports now use simpler cards with each performance metric and its chart together. Shorter labels and fewer repeated explanations keep the focus on rankings, demand and changes. Calculation details remain available from accessible info buttons, while the three charts, heatmap, keyword groups and next steps stay visible by default.
Tracking opens the measured SERP and includes initial keyword data
Keyword details now open the same device and search market as the selected tracker. A measured absence is labelled explicitly, and pending measurements show their progress without prompting another SERP purchase. New tracking setups queue their first measurement together with one-time Essentials enrichment for keywords without basic provider data. Existing data is reused. The setup price includes both jobs; repeat ranking checks do not purchase Essentials again. Both jobs are queued durably, so starting no longer depends on a second browser request. Map clicks and arrow keys now keep the selected area near its current center, including when crossing the date line, instead of shifting longitude by 180 degrees.
Clearer position tracking and geographic checks
Geographic checks stay in the Area tab, leaving the classic reports in place. Map points show available ranking results with a separate indication for incomplete measurements. Tracking orders now offer full settings editing and a choice between Normal and Quick collection, with the selected speed included in the cost preview. First measurements show a compact ranking distribution; missing measurements no longer appear as zero estimated traffic. Measurement details are below the report, and competitor comparisons include a visual summary. The tracking editor scrolls within the window, keeping all settings and save actions reachable on smaller screens. The keyword field starts compact and can be expanded when needed. Initial tracking requests now start both ranking collection and basic keyword-data lookup. Failed starts are shown as failures, and competitor visibility uses the same measured keyword set as the overview instead of treating pending keywords as absent rankings.
Visual ranking reports with comparable checks
Position Tracking is organised into Overview, Keywords, Topics & Pages, Competitors, Search Results and Area. Jobs and settings have a separate entry. Ranking distributions, a keyword heatmap and a competitor matrix turn measured results into navigable analyses. Topic assignments share the existing keyword database tags. The overview highlights keywords near the top ten, lost top-ten positions and landing-page switches. Competitor and SERP-feature trends use a common measured keyword set. Missing observations remain distinct from measured absence. While a new check is running, the last completed check remains available; interim results can be selected explicitly. Keyword details combine saved keyword metrics with ranking history and archived organic result comparisons. Future checks retain organic results across the fetched depth, including available titles and descriptions. Older archives show their actual stored scope. Geographic results remain in Area. Estimated keyword traffic, average position and visibility now have three timelines displayed together, with coloured metric cards to toggle individual charts and a shared 30-day default period. The keyword map links ranking positions and search demand to individual details. Competitor comparisons include recorded AI Overview citations, featured snippets and keywords where competing domains rank ahead. Unknown attribution remains visible. The shared keyword analysis opens inside the keyword database and returns to the same keyword and filters. It uses actual saved mobile or desktop contexts instead of assuming desktop results. The workspace overview now continues loading beyond its first keyword page, with repeated observations counted once per keyword and search context. The keyword view shows the ranking heatmap, keyword groups and detailed table together. Winners, losers, top-three rankings, near-top-ten opportunities and missing measurements are visible without switching views. The position-and-demand chart identifies keywords in its tooltips.
Recover from an interrupted sign-in without an admin crash
Protected pages now wait for the data connection to confirm your sign-in. If that authentication fails or expires, you can retry or sign in again instead of encountering a page crash while loading admin users.
A consistent start for geographic tracking
The Area tab now uses the same introduction layout as the other Position Tracking views. Before setting up a check, you can see which results it provides: a visibility map, top-three and top-ten coverage, and competitors. The start button opens setup, where you review the price before starting a paid check.
Balanced keyword columns and predictable resizing
The Keyword Database now shares available table space across its columns, giving metrics more room alongside keywords. Dragging a column divider starts from its displayed width and keeps neighbouring widths stable. Headers and rows stay aligned, and the narrower resize handle leaves more room to click column headings for sorting.
Edit social editions with clearer controls
Social editions now place Save above the text, keep long posts in a bounded editor and avoid repeated workshop headings. Image instructions survive closing and reopening the image tools. Scheduling can be collapsed again without losing the chosen time, and unsaved changes explain why planning is blocked. Unavailable connections and failed edition refreshes block further delivery until the current state is known. Links to missing editions show an explicit message instead of opening another text. Removing a remote draft is labelled separately from cancelling a publication schedule. On narrow screens, admin navigation opens as a drawer so the article and channel editor can use the available width.
A clearer geographic tracking workflow
Geographic check setup brings the location, radius and devices together beside the map, with the selection and price next to the start button. Results use compact filters and clearer coverage figures, with incomplete measurements kept visible. Choose measurement points on the map or from one compact selector. On small screens, each keyword's result stays readable without scrolling the table sideways. Search result lists open in a dialog, with technical measurement details available on demand. The overview now includes a visibility map and a distribution of rankings and unavailable measurements. Incomplete points remain visibly unknown, even when all usable results rank in the top ten. Invalid measurement contexts are excluded from ranking, page and search-feature summaries while the original evidence remains available. If the map background fails to load, the interface explains how to continue using the other controls.
A clearer keyword database with consistent filters and metrics
The keyword database starts with seven essential columns and four summary metrics. Additional columns and metrics remain available, and saved column choices are preserved. Presets in the existing filter menu help you review unassessed keywords, find rising demand and identify incomplete data. Presets fill in the actual filter controls, which remain available even when their table columns are hidden. The filter button highlights applied filters without adding a toolbar row. Keyword details can expand to use the available workspace. Monthly changes now use the same chronological data in the table, charts and keyword details. The monthly chart shows the latest twelve months, with longer history available separately. Missing values stay distinct from zero. Data explanations identify the market, language, observation period and retrieval dates, and distinguish general market potential and estimated clicks from outcomes for your website. Monetary values without a recorded currency no longer receive an assumed euro symbol. Advanced filters apply consistently to database rows, matching-keyword selection and exports, including language, workflow status, SERP features and additional numeric metrics. List reads respect manual additions and exclusions in hybrid lists. Loading, unavailable metrics and partial summaries are explicit; an empty database, empty list and search without matches each provide an appropriate next step. Applied filters now survive reloads and can be shared through the page URL. Browser Back restores the preceding filter selection. Smart-list creation explains unsupported conditions instead of silently discarding them; saved views retain the complete selection. The table now resolves actual workflow column names instead of falling back to Backlog. Workflow status keeps its meaning after a column rename, and repeated status lookups are reused within each results request. Newly retrieved CPC values with known currency carry that currency into keyword details and table cells.
Browser verification for free tools
Free tools can complete browser verification when the site's security settings previously caused repeated verification errors. Inputs stay in place if a check needs to be retried.
More relevant paths between free checks and guides
Free tools now recommend related checks by the question they answer. The AI Traffic Checker is included alongside crawler-access and answer-structure checks. The homepage links directly to resources, and the owned-data explanation connects to the founder workflow. Overview pages omit recommendations to pages held out of search discovery.
Find and reuse editorial content in one place
Content Studio brings blog articles, glossary entries, newsletters, announcements, ideas and social editions into one admin library, with shared access to the calendar and media library. Open an article to edit the source and its social editions in one detail view. Add an account to generate a distinct channel draft; reopening an existing account keeps its saved version. Tabs can be hidden in the current browser without deleting content. Unsaved text survives tab changes. Each social edition has a shared text editor, account-specific limits and a media preview. Use the AI workshop to describe or dictate changes, review the suggestion, then accept or discard it. Saving and scheduling remain separate actions. Before scheduling, review the selected account, text and publication time and confirm that exact selection. Create images at the article or channel edition. Generation runs on the server, and interrupted uploads resume without generating another image. Image tools and the AI workshop start expanded and can be collapsed again. The media library holds reusable files; legacy LinkedIn and Creative Studio entry points lead into Content Studio. Save the workspace language, writing style, default accounts and timezone in settings. Missing social connections are shown explicitly. Saved drafts and publications are checked against the publishing service; missing remote records retain their history for review. Deletions made directly on a social platform are not detected automatically.
Find the next useful guide while reading
English and German glossary explanations now link to relevant guides and current product workflows near the explanation. The Site Audit article adds contextual paths to audit planning, internal link analysis and recurring monitoring. Existing end-of-article recommendations remain available.
Documentation articles load again
Documentation articles open again when a server error previously prevented them from loading, including the keyword database guides for saved views, metrics, and balance and data refresh.
Find the next useful guide while reading the docs
Documentation guides now connect related tasks directly in the explanation, with an early starting point and further links where a specific question arises. End recommendations follow the article's topic instead of repeating the same keyword links on unrelated pages. The paths cover English, German, French, and Spanish documentation. Product and learning-guide recommendations match the current topic; French and Spanish readers see when a destination is available in English.
Browse documentation by topic
The documentation sitemap groups all guides by topic, with a short description for each page. Jump straight to a topic or open a guide from the directory. The sitemap is available from the documentation menu in English, German, French, and Spanish, and keeps your language when you switch between guides.
Set up keyword monitoring in one resumable workflow
Connected AI assistants can now save a selected keyword list, reuse a matching tracking schedule, and check search-volume coverage through one setup workflow. Authorized tasks can also refresh missing metrics and measure an initial ranking baseline within their existing budgets. The workflow can resume while measurements are processing. Repeating the same request reuses its work. Results show missing measurements, approval requirements, and partial coverage, so a configured schedule is never presented as a measured ranking baseline. Initial selections support up to 100 keywords.
Find the next useful step across tools, features, and guides
Feature explanations and learning guides now link directly to the next relevant check or workflow. The free-tool collection and product assessment join the English and German search surface. Links keep your selected language, and recommendations remain available after the guide. The answer-structure tool now explains more clearly which observations it can make and why they do not predict a featured snippet or AI citation. The homepage, resource hubs and company pages now connect relevant content directly in their explanations. Free checks lead back to the Keyword Database and Site Audit.
Glossary content points to current Crawl Foundry features
Glossary explanations and calls to action now describe current Crawl Foundry features and link to the relevant Keyword Database, Site Audit, or internal link analysis. Outdated product promises and old branding have been corrected across all four content languages. Publication checks now reject known retired product links and legacy brand copy, including later edits to published content.
Reliable AI Crawler Checker reports
The AI Crawler Checker can complete browser verification and display successful reports that previously appeared as unavailable. The English and German checker pages now join the sitemap and are linked directly from Site Audit. Other free tools and glossary pages remain outside this search release.
Complete getting-started guides in translated documentation
The German, French and Spanish Quickstart and Workspace pages use the complete product guides instead of generic summaries. Their page titles and descriptions match the instructions shown on the page. Documentation metadata consistently identifies the site as Crawl Foundry Docs and connects it to the main Crawl Foundry product website.
Reliable keyword updates and enrichment progress
New keywords and completed enrichments refresh the keyword database without changing filters or reloading the page. Recent changes remain available while analytical data is catching up, and failed reads no longer appear as an empty database. Adding a previously deleted keyword now counts its restoration in the confirmation message. Keyword and SERP enrichment progress waits for the saved final result before announcing success. Cancelled keyword enrichments no longer show a success message. Background processing has also been restored after a queue configuration prevented accepted jobs from starting.
Clearer performance charts in Site Audit
Site Audit presents performance results in a more compact layout while keeping the existing metrics, charts and URL details. Color scales place mobile loading time, blocking time and performance scores against their quality thresholds. Distribution bars show the number and share of good, intermediate and poor results. Charts fill the space beside their URL lists, resource breakdowns use complete rows, and the trend axis follows the dates of measured runs.
Site Audit crawl details remain reachable on shorter screens
The Crawls tab now scrolls with the page, keeping crawl history, progress details, and the request log reachable on wide monitors and shorter laptop screens. Opening a running or completed crawl no longer clips the lower sections of the report.
Crawl Foundry links for sign-in and documentation
Website links to sign-in, registration and documentation now default to Crawl Foundry when their destination has not been configured. This also covers older coming-soon settings. Explicitly configured legacy service addresses remain supported.
A consistent Crawl Foundry identity across the website, app and help center
The Crawl Foundry name and logo bring the website, application and help center together. Updated navigation, sign-in pages, browser icons and shared views make the service easier to recognize as you move between them. Sign-in and registration links take you directly to the application. Documentation is available from the navigation, footer and relevant guides, with no coming-soon step. Existing service addresses remain available during the domain transition. Keyword Discovery history and source filters now identify saved query expansion runs consistently in every language. Site setup keeps required and optional fields clearly marked across languages. Blog and glossary articles on Crawl Foundry now identify the current site address as their original URL. Existing article content and the legacy website are preserved.
More room to compare keyword research results
Keyword Research now gives the results table the full workspace width. Compact rows keep keywords and their source labels on one line. Open a keyword to inspect its origins, metrics and saved search results. Research history and frequent terms open when you need them. Clear filter descriptions replace the permanent match tabs, and you can choose which columns to show. Domain, position, trend, traffic and SERP are visible by default when available. Bounded keyword and source columns, clearer column dividers and a compact run header keep more data in view. The keyword column and table headings stay visible as you scroll. Horizontal scrolling is accessible directly above the fixed import controls, including when you are midway through a long result list. Selecting keywords now preserves the table cells and checkbox focus. Unchanged trend charts are reused when you select individual rows or the loaded results. Long result lists now render only the visible rows and a small scrolling buffer. Loading more keywords no longer adds every row and trend chart to the page at once. Selection stays intact while you scroll, and the table headings remain visible. All filter criteria now share one editor. Apply volume, difficulty and other constraints together; typing a search waits for a short pause before refreshing results. Column headings switch consistently between ascending and descending order, with difficulty showing the easiest keywords first. Resetting filters keeps your sorting.
Reliable API requests and authorized recurring tracking
Verification requests and Work Memory filters now accept their documented inputs. Swagger includes the Site Audit configuration body and the correct response codes. The public Swagger console also loads its required JavaScript and styles correctly after deployment. API clients receive retry guidance when a job is rate limited, and duplicate keyword-list names return a validation error. Tracking schedules created or updated through the API or MCP require approved automatic execution and a monthly spending limit. Each recurring run checks the connection and its current budget before reserving EUR Balance. Revoking the connection stops future runs. Existing API-visible schedules without a recorded authorization are paused before their next automatic run. Review and resume them in Crawl Foundry, or reauthorize them through an approved connection with the required scopes and monthly limit.
Open Competitive Research directly in Comparison
Competitive Research opens in Comparison, with a guided start when no comparison exists and saved results when one is available. The navigation now has three tabs: Comparison, Competitors and Statistics. Statistics explains why it needs a comparison. Saved comparison selection, date and incomplete-result notices are available alongside the charts. Existing overview links open Comparison and preserve the selected result. Starting a comparison from the competitor list takes you directly to that comparison.
Set up competitor comparisons with one selection and a clear cost summary
The comparison dialog now uses one field for saved and manually entered competitors. Selected domains appear once, and both analysis options show their explanations before you choose. Your domain suggestions stay closed when the dialog opens and appear when you interact with the field. A separate cost card shows the estimated total, reused domains and remaining balance. Comparisons with insufficient balance stay blocked, while fully reused comparisons can start at no cost. Add a typed domain with Enter or the add button before starting the comparison.
Choose saved domains when starting an analysis
Domain Overview, Keyword Research, Competitive Research, Position Tracking and Site Audit now offer saved domains in an editable dropdown. Select a suggestion or enter another domain. New domain analyses and audits start with the workspace's own website when one is available, and your edits stay in place as workspace data loads. The optional keyword comparison explains its direction beside the field: keywords where the analyzed domain ranks and the comparison domain does not. Competitor selections remain removable, and choosing a suggestion does not start a paid analysis. Long domain lists stay open while you scroll with the mouse wheel or drag the scrollbar, including inside setup dialogs.
Clearer chance scores in keyword details
The chance label in the keyword detail panel now sits below the score ring with clear spacing, keeping it readable without overlapping the chart.
Keyword Database: clearer lists and view selection
The Keyword Database uses consistent page margins for its metrics, toolbar, and content. The list navigation has a plain heading and a clearer selected state, with fewer surrounding borders. List, Matrix, and Kanban now form one view selector. Use the arrow keys to move between views, or Home and End to reach the first and last view. The plus button opens the same dialog used to edit lists. Choose a name, description, icon, and color before creating a list, without shifting the navigation. Smart-list previews load keyword data in small pages. Once loaded, changing rules updates the preview without fetching the workspace again. Smart lists separate their name and appearance from rules and preview in two steps. Back keeps the entered values and loaded preview. Manual lists stay on one step. Keyword detail tabs use the same compact spacing and selected-state colors as the view selector. Keyword details now show saved top results and their capture date in the SERP and competition tabs. The position tab distinguishes missing measurements from an unranked domain and includes traffic potential by target position. Data snapshots start expanded. Saved SERP reconstructions restore AI Overview text delivered in nested response elements, keep results and sources together on wide screens, and hide diagnostic badges without measurements. Late autocomplete responses no longer replace suggestions after switching keywords.
Choose how you discover keywords and keep their origin
Keyword Research now separates keyword ideas, autocomplete and questions, competitor keywords, and AI fanout. Keyword ideas offer a compact or expanded scope with an explanation of the sources and limits before you start. Each keyword can retain several origins. Source details stay available when you save a keyword, discover it again, or export research results. Keywords without measured search volume can also be saved. AI fanout collects search queries disclosed by the selected models during a web search. Results distinguish observed searches, searches whose queries were not disclosed, and failed requests. Original questions and model attribution remain attached to the discovered keywords. The same research methods are available through MCP, including model selection and competitor gaps. Agents can inspect source evidence and reuse saved keywords. The seed picker suggests terms from the Keyword Database and previous research, labels their origins, and offers recent or related saved terms.
Readable volume chart labels in keyword details
Monthly volume charts now adapt date labels to the available width, show the year, and leave more room for volume values. Labels remain separated even when viewing several years of history in the keyword detail panel.
More responsive tables and editor overlays
Tables in Site Audit, project sharing, organization settings, billing, and administration now avoid repeated updates when their data has not changed. Scrolling also causes fewer updates in tabs, keyword previews, and editor highlights. Editor highlights combine rapid content changes into one measurement and cancel pending work when closed. Position Tracking now uses consistent section headings and clearer contrast in its reaction summary.
Request SERP data directly from a keyword
Keywords without saved search results now lead directly to the SERP data request. The lowest-cost option is selected by default; you can choose faster delivery before confirming the price. The empty SERP detail page explains what search results reveal and provides a clear request button. Saved analyses remain accessible, and opening the request does not purchase data.
Set up position tracking with a combined keyword selection
Position Tracking setup now has three steps: domain, keywords, and settings. Your workspace website and saved Site Audit domains are suggested when no tracking domain has been configured, and saved competitors remain part of the ranking comparison without an extra setup step. Add your own keywords directly, then open a list or search the keyword database. One shared selection stays visible while you combine sources; duplicates are merged. Switching sources keeps your selection. Search volumes align in a separate column with space before the scrollbar. Remove individual keywords before reviewing the cost; the schedule tracks that fixed selection within the displayed limit. New schedules default to weekly checks on mobile with Top 10 results. Supported workspace country and language settings are used as starting values. Expand the settings to change them or set a monthly spending limit. Aligned dropdowns keep frequency and depth labels readable; a saved workspace region is offered for location search. The redundant scope panel has been removed. Opening tracking setup from a report with existing rankings no longer stalls the comparison table. Setup also opens directly from a keyword’s result detail. Agents can also create a fixed selection from several list IDs, additional keywords, and exclusions. Weekly, mobile, Top 10 and supported workspace locale defaults apply when omitted; agents must still provide a monthly budget. Single-list schedules retain their live membership behavior. Fixed selections support up to 200 keywords; live-list runs remain limited to 500.
Previously fetched keyword data now appears in columns added later
Keyword Database columns added after an enrichment run now show the data that was already stored for those keywords. Categories are preserved in the table read model, and competition levels are displayed consistently regardless of how the source formatted their casing. Covered data no longer needs to be requested again just because its column was hidden during the original enrichment run.
Choose tracking keywords directly from your Keyword Database
The tracking setup now lets you search and select existing Workspace keywords without creating a list first. Your selection stays visible while you search, and larger databases load in pages. Saved lists and manual entry remain available. MCP growth guidance now connects research to populated keyword lists, targeted data updates, and a measured ranking baseline. Agents are instructed to reuse existing data and schedules, verify saved results, and report unfinished steps. Research-only requests keep their original scope; paid work and recurring tracking still require the corresponding authority. MCP tracking can now follow a saved keyword list. Keywords added or removed from that list apply to the next run, within the existing tracking limits and budget. Fixed selections remain available; existing fixed schedules are not converted automatically.
More reliable navigation in Position Tracking
Switching between Overview and Positions now keeps the tracking data connection and reporting window in place. The Positions table also stays responsive while rankings load and when viewing demo data.
Developer API documentation is now publicly accessible
Developers can open the Crawl Foundry API Explorer and download the OpenAPI contract without signing in. The documentation covers Workspace-scoped stored data, decisions, operations, and approval-aware actions. API access itself is still private. Every request requires a Workspace-scoped service credential and passes through the same authorization, limits, and EUR Balance controls as the Agent Platform. Opening the documentation does not permit anonymous data access or start paid work.
Related keyword suggestions now stop cleanly after a failed request
Related keyword suggestions in the Keyword Database now make at most one automatic request per keyword and language during a browser session. If the request fails or reaches a rate limit, the detail view shows the error and waits for an explicit retry instead of sending the same request repeatedly. Successful suggestions are still reused when the Related tab is opened again.
Clearer pages for owned keyword data and technical SEO
Crawl Foundry now gives data ownership and technical SEO their own focused English and German pages. Teams can see how the free keyword inventory differs from paid data refreshes, and how Site Audit turns crawl findings into prioritized, verifiable technical work. The pages are part of the public search surface and link the underlying workflows to the platform and pricing context without expanding the launch into unrelated content. Until the Crawl Foundry app domain is ready, the existing access buttons remain visible as coming-soon controls without linking to the legacy Crawl Foundry app. Their `/sign-up` and `/sign-in` paths stay fixed, so publishing the app later requires only the new app origin and a redeployment.
Public search discovery now focuses on core product pages
Crawl Foundry's public navigation and search discovery now put the homepage, Platform and Solutions first. Comparison and alternative pages are no longer published, and secondary Tools and Resources pages no longer compete for space in the global navigation. The XML sitemap also excludes legal and utility routes. Blog category, tag and author archives enter search discovery only when at least three published articles support the archive, while each published article remains available on its own URL.
Connected agents can request bounded fresh SEO work
Connected agents can now request approval for fresh keyword research, keyword metrics, competitor research, AI visibility, domain snapshots, and site audits. Each request names the exact capability and includes a maximum charge in EUR. Nothing runs automatically. Access remains tied to the Organization, Workspace, connected client, and owner. Before paid work starts, Crawl Foundry reserves the approved maximum from EUR Balance. It then settles the actual cost or releases the unused amount.
Public website now focuses on English and German
Crawl Foundry's public website now supports English and German. The language switcher and HTML sitemap show only those two languages. Former French and Spanish URLs permanently redirect to the same English path instead of returning errors or remaining as separate pages. Existing translation data remains in the codebase. French or Spanish will return publicly only when Crawl Foundry can maintain the product experience and content quality in that language.
Public pages now declare their language in the HTML
Crawl Foundry's public pages now declare the active language in the server-rendered HTML. English pages use `en`, and German pages use `de`. Browsers, assistive technology, and crawlers can now identify localized content before client-side code runs. Canonical URLs and language alternatives remain unchanged.
Connected agents can check their approved access reliably
Connected agents can now list the authorization grants for their selected Workspace even when no grant exists yet. An empty list is returned as a valid state instead of interrupting the connection. The request remains bound to the Organization, Workspace, connected client, and owner. Read limits still apply, while checking grant status does not reserve EUR Balance or start paid SEO work.
Core SEO workspaces now share one clearer reading order
Dashboard, Action Center, Domain Overview, Keyword Research, Keyword Database, Position Tracking, Site Audit, and Activity now follow the same reading order: current state, supporting evidence or change, the next useful action, and how to verify the result. The underlying metrics, charts, filters, tables, and specialist views remain available. Long reports use the page as their vertical reading flow instead of splitting the work across several small scroll areas. Wide tables and module tabs remain horizontally reachable on narrow screens. Loading, empty, partial, stale, failed, and not-yet-measured states stay distinct where the available evidence supports them. Action Center and Position Tracking also keep their relevant view, scope, filter, and detail context in the URL. Shared links and browser navigation can therefore return to the same working context instead of reopening a generic overview.
Each product page now names the SEO tool it provides
The homepage now presents Crawl Foundry as on-demand SEO software that does not require a subscription. It explains the free organizing and export workflow, then separates that work from fresh data requested only when needed. Platform, Keyword Database, Site Audit, and the seven feature pages now name their own product category in the main heading and search metadata. Decorative sample queries in the homepage hero are painted visually instead of appearing as page-copy text nodes. Each homepage chapter now links to the relevant product or solution page. A new evidence section also explains paid-action scope, data freshness, and the limits of later comparisons. Localized pages set the active document language without giving up static rendering.
Workspace analyses stay compact and show the data already available
Domain Overview and Competitive Research now use the available viewport height consistently. Competitor lists stop growing after their maximum height and scroll inside their panel, while the associated maps and charts use the remaining space. Site Audit now keeps one intentional content scroll region across its dense views. Click-depth and issue-category summaries are more compact, labels remain readable, and charts no longer force adjacent cards to stretch. Position Tracking shows saved ranking history even when the newest scheduled run could not start, and reports that run status accurately. Starting Keyword Research keeps the active state lightweight instead of loading result datasets before they exist.
Main marketing pages now state their search intent clearly
The homepage now identifies Crawl Foundry as an AI-assisted SEO platform for keyword research, technical Site Audits, prioritization, and measurable SEO workflows. This category statement appears in the first visible section as well as the page metadata. Platform, Features, Keyword Database, Site Audit, Solutions, Pricing, Resources, Learning Center, Tools, and Owned Data now have distinct titles and descriptions that match each page's job. The shorter descriptions improve snippet clarity without promising rankings or traffic.
Agent connections now start with the work, not the protocol
The Agent Platform now opens with a guided MCP connection for Codex, Claude, Visual Studio Code, Cursor, and other compatible clients. The selected workspace is built into the connection, OAuth is the standard path, and the technical endpoint and Inspector remain available only when they are needed. The capability view now explains 21 focused SEO tools instead of presenting the much larger internal operation catalog. Keyword research, domain and competitor analysis, Site Audit, tracking, planning, content, publishing, and verification are grouped by the result an agent can deliver. The Developer API has its own advanced area with workspace-scoped access, billing guidance, and an API Explorer. MCP and REST use the same organization EUR Balance, approval policy, spending limits, and settlement controls.
Site Audit examples are now clearly labeled
The public Site Audit overview now identifies its illustrated crawl counts and run history as sample data. The preview remains concrete without looking like a live customer audit. Each example issue family also explains its own impact and verification path. Broken links, missing canonicals, redirect chains, and orphan pages no longer share one generic canonical explanation, and the displayed URLs are explicitly presented as examples.
Billing now separates cash, balance, invoices, and receipts
Billing history now answers two different questions separately: how much cash was paid or refunded, and how much prepaid EUR balance was added or removed. That distinction matters when a plan includes more balance than its purchase price. Older transactions whose paid amount cannot be reconstructed are marked as unavailable instead of being presented as estimates. Completed Polar orders expose the invoice or payment-receipt links Polar returns. If a document link is unavailable, Crawl Foundry points to the Polar customer portal rather than creating a second invoice or inventing a receipt. Portal access remains available after a downgrade to Free, so earlier orders, billing addresses, tax details, and payment methods can still be managed there. Before a new subscription or top-up, the buyer identifies the purchase as private or business use. Polar remains the merchant of record and collects the billing and tax details in its checkout. The confirmation explains that Crawl Foundry does not offer voluntary refunds for unused subscription time or prepaid balance. Mandatory rights and refunds initiated by Polar remain unaffected. Public cancellation and consumer-withdrawal routes give customers a direct path outside the product navigation. The cancellation form distinguishes ordinary from extraordinary cancellation and asks for the important reason only in the latter case. Both declarations produce an immediate reference and an email confirmation. Signed-in customers can reach Polar cancellation without a feedback survey in between. The public forms and their delivery were browser-verified. Authenticated checkout, document, portal, renewal, refund, cancellation, and withdrawal paths still require controlled production transactions before the entire provider flow can be called live-proven. This release is not a claim of legal or tax adviser approval.
Keyword discovery finds more, remembers paid data, and reaches back to 2019
Deep keyword discovery now combines two complementary expansion methods. A category-based lookup covers all seeds in one request and returns volume, CPC, difficulty, intent, and a 12-month trend with every keyword. Imported candidates therefore arrive with the base metrics already collected. A phrase scan adds up to 300 ideas per seed that contain the seed wording directly. Both methods appear as separate items in the cost preview before a run starts. Discovery also reuses paid data more consistently. Keywords that already carry base metrics from an earlier discovery import count as covered during freshness checks and automatic enrichment after import. The same base dataset is not purchased a second time simply because the keyword entered the workspace through a different route. The keyword dossier adds an optional "History since 2019" package. It extends the 12-month trend into a multi-year monthly series and was introduced at about one cent per keyword, passed through at source cost. The package can be requested for one keyword from its trend section or for a selection in the enrichment dialog. Every discovery row now opens a dossier. A keyword already in the database opens its full workbench view. A candidate that has not been imported opens a compact read-only view with the metrics purchased by the current run, so the evidence is available before the import decision. Automatic base-data enrichment after an import now uses a lower-cost delivery option when the primary source has no record for a keyword. This route costs about one third less than an immediate request and normally returns volume data within a few hours. Interactive enrichment keeps the immediate option when someone is waiting for the result.
The competitor catalog now connects paid data with practical decisions
Each competitor now opens with an overview of its organic footprint: ranking keywords, estimated traffic and value, recent movement, and position distribution. Ranking tables retain the exact URL behind each keyword, while source badges explain why a domain entered the catalog and which query last surfaced it. Existing Domain Overview evidence is reused when available instead of prompting for the same purchase again. Competitive Research now separates overview, catalog, comparison, and statistics. Comparisons show module coverage, collection dates, failures, and a verdict with the proposed next step; incomplete runs open their evidence instead of looking complete. Deeper data loads only for the tab being viewed, which reduces the initial load while keeping traffic history, keyword profiles, link signals, and the market map available when needed.
Domain Overview turns more stored evidence into working views
Domain Overview now turns more stored analysis evidence into working views. Rankings include SERP-feature coverage, volume trends, and position charts; competitors show their ranking profile; pages show traffic concentration; and backlinks add authority, link-position, and referring-domain distributions. History separates estimated traffic growth, keyword movement, ranking shares, and churn. Each chart states its basis and uses bounded server aggregates rather than only the rows visible in the current table. The new Opportunities view identifies promising ranking ranges with a disclosed CTR estimate, brand heuristics, SERP-feature exposure, and relative authority. These are decision signals, not traffic promises or causal claims. Keywords can move into a list or research, competitors can be saved, and Domain Overview can cross-check important pages and linked targets against the latest Site Audit. Pages outside that crawl remain unmatched rather than being labelled healthy.
Sort the Domain Overview rows you have loaded
Domain Overview tables now follow the same interaction pattern. Rows highlight on hover, headers stay visible while you scroll, and a click on a data header sorts the results without adding a separate control bar. Results still arrive in small server pages. Sorting applies only to the rows currently loaded and recalculates when more rows arrive, so it should not be read as a global sort across the full analysis. Top Organic Keywords can now also sort by ranking URL and search intent. Keyword Gap, the complete competitor table, and the Rankings, Pages, Competitors, and Backlinks tabs use the same behavior.
Keyword imports can add missing base data after showing the price
The default pricing policy for keyword enrichment now passes the data cost through without a markup. The quote shown before a run remains the source of truth if that pricing policy changes. It breaks the selected data packages and fallback routes into a visible estimate before any paid work starts. When research results are imported, enrichment is selected by default for the keywords still missing base data. The import dialog estimates that missing scope and lets you turn enrichment off. Keywords that already carry fresh paid metrics are not purchased again. An optional long-tail fallback can ask a second volume source about terms the primary database does not cover, although that source may still return no data. The 12-month histories already returned during research are now stored in the Keyword Database instead of being discarded. When that history is available, trend sparklines can appear as soon as the import finishes.
Activity filters now reconcile the whole view and its export
Filtering Activity by area, status, initiator, cost state, or calendar day now changes the whole view, not just the timeline. Headline figures, the daily chart, the atlas, the category breakdown, and spend per member all recalculate for the same structured scope. Free-text search still refines only the entries already loaded, and the interface labels that narrower behavior. The activity calendar now covers twelve months independently of the shorter analysis window. Daily rollups provide exact calendar totals rather than a sample, with the current day recalculated from live operations. While older history is still being backfilled, the view marks that state and uses a bounded fallback instead of pretending the rollup is complete. Every view can be exported as a CSV file containing the rows currently visible. The file records its own scope, including the window, row count, and creation date. Cost columns are included only for teammates with billing permission. Optimizer and Precision Editor runs recorded after this release now appear as named operations with their own outcome and duration, rather than unattributed usage costs. Earlier runs were not reconstructed retroactively.
Large Site Audit reports no longer reread every link during publication
Current crawls keep exact link counters with each URL. Report publication now reuses those counters instead of loading thousands of link records again. Only older crawls without counters still need that extra read. This avoids the cumulative read limit that could leave a crawl complete while its report stayed unfinished. A completed or partial run remains eligible for automatic recovery until the required publication stages finish. If publication stalls, or report totals are still being calculated, recovery can continue after the run has been idle long enough. There is no age cutoff for this incomplete publication state.
Activity now explains workspace work and settled spend
Activity now brings the work from across your SEO workspace into one history: Domain Overview, Competitive Research, keyword work, Position Tracking, Site Audit, Search Performance, Action Center, Research, Content, and Brain. A workflow appears as one understandable operation, rather than a trail of technical background steps. The redesigned view adds daily activity and spend trends, a calendar heatmap, category development, attention states, source coverage, filters, and an expandable timeline with links back to the product where the work happened. Settled spend can also be broken down by the workspace member who initiated the billed work, including each member's share and daily trend. Only teammates with billing permission receive that breakdown. Everyone else keeps the operational history without cost or contributor data. Selecting a calendar day filters the timeline, and the control for clearing the selection sits beside that date in the calendar header. The timeline covers the latest 100 normalized operations. If a source or reading cap makes a figure partial, the coverage view says so instead of presenting it as a complete total.
Site Audit website setup now follows the current crawl limits
Adding a website to Site Audit no longer fails when an older browser default exceeds the crawl limits currently available to the workspace. New sites now inherit the active limits used when a crawl starts, while existing settings remain compatible. Crawl-depth and URL-count fields can be cleared and entered again without snapping back to their previous value. Their maximum follows the active product limit rather than the site's current selection. Existing website settings now expose the same crawl scope as initial setup, including depth, included and excluded paths, rendering, and robots.txt handling. Saved changes apply to future runs. New audits also recover when the dedicated crawl process loses its connection during startup. Instead of leaving a run at the first seed URL while appearing healthy, Site Audit retries the connection and restarts the failed process if recovery does not succeed within its limit. Monthly run allowances now treat no-result outcomes consistently. Cancelled, failed, and partial crawls return their run slot when fewer than ten pages completed. Completed crawls, and incomplete crawls that produced meaningful page work, still count. Page-based billing remains tied to the pages actually consumed.
AI Crawler Checker now opens every check as a full report
AI Crawler Checker results now open in a dedicated report instead of appearing below the form. Four eligibility verdicts lead the page, crawler rules are grouped by provider, and fetch, indexability, and extraction details sit in a collapsible evidence section. The layout uses the full screen on mobile and records exactly when the page was checked. Repeated checks of the same URL can use a result stored for up to 30 days, making shared and revisited reports available immediately. A "Run a fresh check" action requests the live page again under the same verification and usage limits. The extraction measure now compares visible text inside the main content instead of comparing text with raw HTML bytes. That avoids penalizing framework-heavy pages simply because their markup is large. Internal-link evidence also focuses on links within the main content and quotes readable anchor text rather than markup fragments.
The robots.txt checker puts crawler decisions in one report
Results now open in a dedicated report instead of below the form. The report begins with the number of search, training, and user-fetch crawlers allowed for the checked path, then shows the winning directive and source line for each supported bot, grouped by provider. It remains readable on a phone without horizontal scrolling. Successful checks are cached for up to 30 days. Definitive errors are kept only briefly, and network failures are not cached. The report shows when the file was fetched and whether the result came from the cache. After a robots.txt change, "Re-check now" bypasses the stored result and reads the live file.
The AI Traffic Checker separates local analysis from URL checks
Running an analysis now opens a full-screen report with AI sessions, sources, recognized rows, and broken destinations at the top. Sections for sources, landing pages, URL status, and data quality follow in one scroll. The drop zone maps common GA4, Plausible, and Matomo columns and asks for help when a required column is missing. CSV rows are analyzed in the browser and are not uploaded. Only the deduplicated URLs selected for status checking are sent to the protected checker. Status results are cached for 30 days and show their age. A manual re-check bypasses the stored result and counts against the existing daily limit. Browser verification and a silent trap field reject automated submissions before they reach the checker.
Brand Voice Analyzer can profile a public page by URL
URL mode now fetches a public page, extracts its main text, and analyzes that text in the browser. When the page can be fetched, sentence rhythm, perspective, style markers, recurring phrases, and writing guidelines open in a focused full-screen report. Successful fetches are cached for 30 days, and the report shows the timestamp and cache state. Definitive fetch errors are cached for five minutes; transient failures are tried again on the next check. The extracted sample is capped at 100,000 characters, and URL checks are protected by browser verification. Paste mode makes no server request: its sample goes directly to local analysis in the browser.
Featured Snippet Checker can now inspect a live page by URL
The free Featured Snippet Checker can now fetch a public page directly. Paste a URL and the report extracts the main content, reviews answer structure and heading hierarchy, and shows whether important passages can stand on their own. The result opens in a focused full-screen view instead of a narrow column beneath the form. Checked pages can be reused from a snapshot stored for up to 30 days. The report shows when that snapshot was taken, so an instant repeat result is never confused with a new fetch. A short browser verification protects the public URL check from automated abuse. Pasted-text analysis remains local in the browser.
Crawl Foundry's free checks are easier to find where they help
The six free checks now appear next to the work they support. Site Audit points to AI Crawler Checker, Keyword Discovery to the People Also Ask Tool, SEO Monitoring to AI Traffic Checker, and the technical SEO solution page to the robots.txt checker. The homepage introduces the free-tool family, and the platform page links to the complete overview. Each recommendation says which question the tool answers and makes clear that no account is required. People can move from a product concept to a bounded public check without first searching the site for the right utility.
AI Crawler Checker shows what AI crawlers can actually access
The free AI Crawler Checker now examines one public URL and shows what a crawler can actually reach. It follows valid public redirects, checks the final site's robots.txt, and reports page access, indexability, canonical and snippet controls, crawler-specific rules, extractable content, heading structure, and internal-link context. Every verdict comes with the evidence behind it. The report includes redirect steps and relevant X-Robots-Tag directives instead of reducing a complicated setup to a vague yes or no. Compressed responses are handled within the same fixed check limits, while unsafe or private destinations remain out of scope. The checker does not execute page JavaScript or send the inspected content to an external text model. That boundary is visible in the report, so teams can understand both what was checked and what still needs a browser or manual review.
AI Traffic Checker now verifies referral URLs
The free AI Traffic Checker can now verify a bounded list of public referral URLs. For every address, it reports whether the page responds, where redirects end, which final URL was reached, and why a check failed. Each URL is checked independently. If one source is unavailable, the other results remain usable and the failed check stays visible instead of being folded into a misleading success state. Public-address validation, fixed response limits, traffic controls, and timeouts keep the free check predictable without hiding partial results.
AI robots.txt Checker explains crawler access rule by rule
The free AI robots.txt Checker now reads one public site's robots.txt and explains how its declared rules affect known AI crawler agents. It distinguishes allowed, blocked, and conflicting cases instead of turning the entire file into one unsupported verdict. For each crawler, the result shows the matching directive and its source line. That makes broad user-agent rules, more specific overrides, and contradictions easier to review. The check accepts only public destinations and uses fixed content, time, and request limits so a malformed or unusually large file cannot turn a quick diagnostic into an open-ended operation.
Brand Voice Analyzer now builds a complete private style profile
The free Brand Voice Analyzer now turns a 100-word-or-longer writing sample into transparent sentence, paragraph, punctuation, perspective, marker, repetition, and lexical-variety evidence in English, German, French, or Spanish. Samples from 300 words receive the full profile; shorter accepted samples are clearly marked as limited. The complete analysis stays in the browser, so even a sample near the 100,000-character limit does not tie up the interface. Submitted text is never uploaded, logged, or sent to an external text model. The resulting brief can be copied, and the structured profile can be downloaded as JSON without including the original sample.
Featured Snippet Checker handles large and ambiguous documents more clearly
The browser-based Featured Snippet Checker now skips empty structures when it looks for a direct answer. It reports the length of the selected passage and points out dependencies that make an answer difficult to understand without surrounding context. Large documents now enter a visible analysis state before local processing begins. Findings, evidence, outlines, detected structures, and question sections appear as separate previews. The on-page result remains bounded even when a document repeats thousands of headings, while Markdown, checklist, and JSON exports retain the complete set of findings.
People Also Ask Tool separates observed suggestions from assumptions
The free People Also Ask Tool now collects a bounded set of observed Google Autocomplete suggestions for one topic, language, and market. It groups the returned ideas into questions, comparisons, alternatives or reviews, and utility searches, and shows which source strategy produced each suggestion. The report distinguishes a live observation from a previously stored result. It does not present autocomplete suggestions as a copied People Also Ask box, search volume, ranking evidence, or frequency data. Clear empty and partial states show when the source returned less than expected, while CSV and JSON exports preserve what was actually observed. A full reset makes repeated research runs easier to compare.
Six free SEO tools first opened with honest preview pages
On July 30, the free tools hub opened six pages covering crawler access, question discovery, answer structure, writing voice, and AI referrals. At that release, each page explained the input, evidence, method, an illustrative report, and where the result could lead inside Crawl Foundry. Those first pages did not run an analysis. The previous SERP Analyzer form, which never produced a report, redirected to the hub instead. The preview state was temporary: later releases moved each tool to a live engine once processing, abuse protection, limits, and result validation were ready. The hub now marks all six tools as live.
Action Center saves calibration answers against the current profile
When a business-fit question comes from an older Action Center report, choosing "relevant" or "not relevant" now applies the answer to the current business profile. Existing calibration choices are kept, so a profile refresh no longer turns the next answer into a revision conflict. The buttons wait until the current profile is available and lock while a choice is being saved. This rebasing applies only to calibration answers. A direct edit to an outdated profile is still rejected rather than overwriting newer work.
Product limits now use one server-side policy catalog
Dashboard, Action Center, Domain Overview, Keyword Research, Competitive Research, Position Tracking, Site Audit, Activity, and the Keyword Database now use stable definitions from one policy catalog for the boundaries connected to it. Product, plan, and safety limits remain separate, and each eligible request resolves one coherent policy snapshot. A lower connected limit applies to the next eligible request; guidance shown in the interface is not the final authority. Limit checks happen before paid work starts. If a request exceeds its current allowance, Crawl Foundry keeps the input available, starts no paid run, and reserves no balance. Not every definition is editable: deployment-only and diagnostic values stay fixed, while definitions that do not yet control a verified product boundary remain read-only.
Support replies stay in the original email conversation
Each bug report or help request now has a ticket reference linked to one email conversation. When the Crawl Foundry team replies, the message uses the existing conversation and standard email reply headers, so separate messages do not have to be matched by hand. A customer reply is attached only when its conversation or ticket reference and sender can be verified. A verified reply reopens a resolved ticket and returns it to Needs reply; duplicates are ignored. The team must reopen a resolved ticket before sending another response.
Competitor metrics refresh after stored SERP runs
Shared keywords, covered volume, average position, and keyword gap used to change only when someone selected "Recompute" in the competitor view. They now refresh automatically after a tracking or Keyword Database SERP run stores new results. The calculation uses the workspace's stored data, so it does not make another paid request or reserve balance. A newly catalogued domain still needs to appear in stored SERP results before those figures exist. Until a stored run records it, the four metric cells show a dash and explain why no value is available. A numerical zero now represents a measured result rather than missing data.
Deep research now has a larger result budget than Quick
On July 27, a Deep run could return fewer usable keywords than a Quick run from the same seeds. Its expansion was narrower, and some suggestions arrived without the metrics required for import. This release widened the Deep pass, added bounded follow-up expansion, and checked the strongest candidates for metrics. The recipe has been refined since then. Deep currently has a larger source and result budget than Quick, with a cap of 2,000 deduplicated rows compared with 600. These are ceilings, not promised yields. Coverage depends on the available data, and a row without metrics remains visible but cannot be imported. The estimate shown before the start includes the paid steps selected for the run.
Keyword research match modes now show their results
Switching a research run to Phrase, Broad, Exact, Related, or Questions used to filter only the rows that happened to be in the first result window. A valid match farther down the run stayed invisible until someone clicked "load more" several times, which made a populated run look empty. Match modes now look farther into the run, so selective matches can appear on the first page while large runs keep a bounded search window. The workbench also requests the next batch while a filtered result is still incomplete. The filter bar shows how many candidates match the active mode, and the empty state distinguishes between three different situations: the search is still running, no match has been found yet, or the available candidates have all been checked.
Position Tracking shows live measurements and recovers stalled runs
A measurement in progress no longer looks like a failed one. The run log, run-health chart, and a live band on the Position Tracking page update as results arrive, for example "2 of 4 loaded." Missing results are counted only after the run has reached a terminal state. Until then, the interface reports coverage rather than drawing a completed red bar from unfinished work. Pressing "run now" while a healthy run is active does not replace it. If the click starts nothing, the response names the reason: another measurement is running, the tracker is not active, EUR Balance is insufficient, or the monthly budget cap has been reached. The interface no longer confirms a start when no run was created. SERP trackers also expose a separate fast action. It sends that single manual measurement through the priority queue and shows its estimated EUR cost before the click. Scheduled measurements keep their standard queue and price. A tracker configured for desktop and mobile dispatches both device runs as one group, and the progress band reports their combined coverage instead of showing two unrelated half-runs. If processing stops reporting after result requests have already been paid for, the run is no longer abandoned indefinitely. Recovery reuses the registered requests, collects any results that became available, and writes them through the same per-keyword settlement path as the original run. Results are not ordered or charged a second time. A fully recovered run succeeds; incomplete evidence stays partial, and an unrecoverable run ends as failed with a stated reason instead of remaining active forever.
Internal linking gets an actionable score
The old internal-linking card led with an average percentile that was structurally 50 and two correlation coefficients. Those figures described the calculation, but they did not tell you whether the site's link structure needed work. The card now reports one score from 0 to 100 and one of four verdicts: Strong, Solid, Improvable, or Weak. A short state sentence explains the result and the next sentence names the weakest part to address. Three scored aspects show how the verdict was formed: whether important pages are reachable at a sensible depth, whether internal authority is distributed without becoming flat or overly concentrated, and whether internal-link density follows the hierarchy. The raw correlations and concentration share remain in the method tooltip for readers who need to inspect the calculation.
Legal pages match the website and fill content gaps
Privacy, Terms, Cookies, and Imprint now use the same navigation, branding, and footer as the rest of the public website. The reading layout adds clear section boundaries, restrained line lengths, and tables that remain usable on a phone. A document switcher links the four pages without sending readers back through the footer, and it keeps the selected language. The imprint now names the current operator with a postal address, VAT ID, and contact address. It also distinguishes that operator from Contextter UG (haftungsbeschränkt) i. G., which is still being incorporated, and replaces the outdated media-law reference that appeared before this update. Several gaps in the legal copy were closed at the same time. The Terms explain EUR Balance from plans, top-ups, and trial balance instead of rendering empty list items. The Privacy Policy identifies the controller, gives the stated legal basis for each processing category, and explains the right to complain to a supervisory authority. The Cookie Policy separates the signed-in application's cookies and product analytics from the public website and help center, whose standard measurement setup does not use analytics cookies. German, French, and Spanish versions carry the same sections rather than falling back to incomplete English fragments.
Pricing copy in every language and prerendered marketing pages
Plan names, descriptions, feature lists, and calls to action on the pricing page now use the selected German, French, or Spanish catalog instead of silently falling back to English. Prices and plan limits still come from the shared plan configuration, so localization changes the explanation rather than creating a second pricing source of truth. Most marketing routes are prerendered again. They no longer wait for request-time data they do not use, which removes unnecessary server work from each visit. Blog and glossary routes remain dynamic where preview and filtering require fresh data.
Site Audit explains what it keeps and how crawl quotas count
Site Audit crawl history now has an explicit lifecycle. Recent runs and at least the ten newest terminal runs keep page-level evidence for comparisons and fix verification. Before an older run is compacted, its expiry is announced. Compaction removes heavy page detail but keeps the summary and the evidence used for health history and trend charts. Active comparisons or exports postpone that cleanup. After roughly six months, one completed run per calendar month remains as a compact long-term anchor; if a month has no completed run, its latest partial run can fill that role. Other runs in the same period can be removed. Runs past twelve months are removed except for the site's newest terminal result. Up to five runs per site can be pinned against earlier compaction, but a pin does not override the twelve-month boundary. An archived site has a seven-day recovery window before its audit history is deleted. Crawl quotas are separate from URL limits. The current standard plan values are fifteen full crawls and ten probe runs per organization and calendar month; the effective values can differ when plan or administrative limits are changed. The start dialog shows the applicable allowance, remaining runs, and reset date. Each crawl still uses the URL ceiling of its plan. Active and completed full crawls consume a quota slot. A cancelled, failed, or partial run with fewer than ten completed pages does not, and a run skipped for insufficient EUR Balance never started and is not counted. This keeps a failed start from using the same allowance as a crawl that actually processed the site.
Breadcrumbs show where each public page belongs
Public product, solution, tool, legal, blog, and glossary pages now show where they sit in the site. The breadcrumb leads from the home page through the relevant section to the current page, and every intermediate level is a working link. Related topics keep their product context: keyword research and tracking sit under the Keyword Database, while technical reports sit under Site Audit. The same hierarchy is included as BreadcrumbList structured data. Search engines can use it to understand the page relationship and may show that path in a result, but the presentation in search remains their decision.
One contact address across the public website
Public contact links now point to info@crawlfoundry.com. That includes the company and legal pages, privacy information, structured company data, contact forms, cancellation and withdrawal fallbacks, and newsletter error messages. All of these surfaces now use the same address.
Domain Overview puts ranking movement in context
Domain Overview now shows ranking distribution as a stacked history. The chart follows the same domain through the Top 3, positions 4 to 10, 11 to 20, 21 to 50, and 51 to 100; its upper edge is the total number of ranked keywords. That makes a growing total distinguishable from a real move into stronger positions. Search intent is calculated across the persisted ranked-keyword set, not only the rows currently loaded in the table. Each intent shows its share, keyword count, and estimated traffic. Competitors appear on a logarithmic positioning map with ranked keywords on one axis, estimated traffic on the other, and shared keywords controlling bubble size. Source context stays quiet unless a section genuinely uses a different data state from the current profile.
Action Center groups internal-link suggestions by target page
Internal-link recommendations became one entry per target page. Previously, each stored link suggestion created another row, allowing the same page to appear several times in the recommended order. The grouped entry collected the suggestions for that target. The figure beside the entry became the count of bundled link suggestions. It no longer displayed an internal quality score under a label such as “69 concrete link recommendations”. Recommendations also replaced storage-oriented wording with a decision the user could verify. A recorded position 9, for example, became a prompt to review the ranking page, search result, intent, and available on-page evidence before choosing an optimization. It was no longer presented as proof that the page would climb. The page title also appeared only once.
Workspace Home adds a visual ranking summary
The **Your rankings** card began drawing the average position of tracked keywords over time and split the latest recorded positions into Top 3, Top 10, Top 20, and Top 100 bands. The chart direction made improvement and decline visible without changing the underlying position values. A short footer stated what was measured, how often the schedule ran, and when it last completed. The card replaced two text-only panels. The keyword trend chart also named its subject, **Visibility of your domain**. Its Keywords, Traffic, and Top 10 series described the domain-level organic evidence rather than the selected list or the separately tracked keyword set. The activity feed moved out of the main area but remained one click away in a compact card. The full log stayed on Activity, and the dashboard help tour followed the revised layout.
Domain Overview adds controls for loaded evidence
Domain Overview began allowing the loaded competitors, ranking keywords, top pages, anchors, referring domains, and linked pages to be searched and sorted without starting another paid analysis. Position groups filtered the keyword evidence directly, while competitor rows and the traffic comparison highlighted the same selected domain. The growth charts shared a 6, 12, 24, or 60-month window. Active controls stayed in the page URL, allowing a focused view to survive a refresh or be shared. Every result counter stated that it referred to loaded evidence, keeping a partial snapshot or paginated result identifiable.
Keyword Research closes runs with expired balance reservations
Keyword Research began closing a run with an expired EUR Balance reservation through the same visible failure path as a processing or handoff failure. The unused reservation was released, the run received a support code and activity record, and another research request could start without stale capacity from the earlier run. Timeouts, inactive reservations, stalled processing, cancellations, and authorized recovery retained distinct diagnostic codes instead of collapsing into a generic internal error. Customers saw the stable explanation and support code; operational detail remained available only to authorized operators.
Site Audit publishes retained link evidence with visible progress
Site Audit began evaluating contextual internal links and repeated anchor text from the retained link graph rather than a fixed first-page sample. Large crawls therefore kept findings beyond the first 1,000 matching links whenever the retained graph was complete for that rule. The processing indicator reserved a measured final-publication stage and named the work in progress, such as link metrics, issue summaries, aggregates, or the Health Score. It did not move to 99% before that report work had begun. Audits also identified identical title and H1 text, external nofollow links, and unnamed external links as review diagnostics. Valid decorative images, accessible image links, intentional resource links, and an absent `llms.txt` remained outside the defect count. Empty image alternatives became visible without being confused with a missing attribute. The audit distinguished likely content images from explicitly decorative markup and kept both outside the Health Score until review. Issue exports showed known categories, evidence sources, and target URLs for duplicate anchors instead of generic labels. Repeated template patterns used a concise, representative URL sample.
Site Audit validates structured data and reports measured performance
Site Audit began validating JSON-LD, Microdata, and RDFa beyond syntax. It checks Schema.org types and properties, basic value formats, deprecated vocabulary, and, for supported Google result types, the difference between required and recommended fields. The overview separates clean, review, blocked, and unmarked URLs, groups repeated template problems by cause, and keeps the item-level evidence behind every verdict. History and URL filters make it possible to follow a type or result feature across crawls without treating a sample as full coverage. Performance reporting now keeps crawl response time, lab measurements, and available field data separate. Every run states how many eligible URLs were planned, measured, or failed. Missing CrUX data stays unavailable, and a slow Lighthouse measurement can reduce performance coverage without failing the completed crawl. Measurement profiles are versioned, interrupted work can continue, and partial coverage remains explicit in exports and comparisons.
Action Center shows one recommendation at a time
The Action Center was redesigned around one practical question: what should you review next? Instead of opening with a dense multi-column dashboard, it showed one recommendation with a plain title, the reason for its current position, and the three headline measures behind it. From that card, the recommendation could be started, saved for later, or marked as not relevant. The last choice became an explicit business-fit signal rather than silently changing the shared guide. Step-by-step guidance, the supporting evidence, ranking feasibility, and optional data enrichment stayed available in the detail view. A grouped list kept the broader guide in **Now**, **Up next**, and **Watch** bands, with in-progress items still reachable. After the user had reviewed the current set, “done for today” described the state of that review queue, not the health or completeness of the website.
Competitive Research keeps paid comparisons recoverable
Competitive Research began reusing the same comparison when a browser or network retry repeated the request. The balance reservation stayed attached to that operation, allowing an interrupted handoff to recover without creating a second paid run. Full competitor analyses published one complete ranking generation at a time. If source retrieval, score calculation, or finalization failed, the previous successful data remained visible and unused reservations were released. A partly written result did not replace the last usable comparison. Comparison reports distinguished running, partial, failed, empty, limited, and available modules for each domain. Coverage, freshness, applicable limits, and support codes stayed visible. The report also added decision guidance, detail dossiers loaded on demand, stable shareable URLs, paginated history, and a CSV export that preserved missing data instead of turning it into zero. Quick comparisons marked which modules required a Deep analysis and offered a prefilled upgrade. If Quick and Deep started close together, they reused the same running overview data instead of reserving and requesting it twice. Deep comparisons consolidated already purchased keyword-gap evidence across the selected competitors. The bounded overview prioritized gaps shared by the most domains, kept unavailable metrics visibly unknown, disclosed sampling, and handed individual or grouped opportunities into Keyword Research without a new source request.
Workspace Home shows measured values instead of setup prompts
The average-position and opportunity figures on Workspace Home began using the workspace's recorded evidence. Average position read the active rank tracker, including a tracker with hand-picked keywords. The opportunity count used the same workspace-wide evaluation as the Action Center. A measured zero appeared as zero rather than as a prompt to configure work that was already running. Every point in the keyword trend received a larger hover target. Its tooltip stayed inside the visible card, and a legend named both lines when the Top 10 view also drew the Top 3 series. The chart stopped creating a separate small scroll area inside the card. The tracking panel added a compact trend for the primary schedule's average position beside its keyword count. Once every setup step was complete, the checklist collapsed to one confirmation line. The competitor card explained that a blank marker meant "not measured yet" rather than a broken result. At this release, new workspaces also saw what their available starting balance covered and where they could add funds.
Domain analyses separate missing data from an empty result
Domain analyses began stating the direction of every keyword gap: the first domain ranked while the comparison domain did not. When the user analyzed a competitor, the workspace's tracked domain became the default comparison. Analysis of the workspace's own domain offered its tracked competitors and selected the only one automatically when there was no real choice to make. Each data section kept its own retrieval state and coverage. A failed source request appeared as missing evidence rather than as a genuine zero-result finding. Limited samples showed how many rows were loaded against the reported total. Analysis time remained separate from source freshness when a freshness timestamp existed. Starting another analysis from a saved snapshot preserved its market, language, keyword scope, and comparison domain. Already purchased fields were retained more completely, including shared and full competitor metrics, page-level organic metrics, keyword-gap landing pages and traffic, backlink spam and crawl facts, and indexed-page totals. The history marked complete, partial, running, and failed attempts directly. It opened the latest usable result instead of allowing a newer failed attempt to hide it, and showed a loading state while older keyword, competitor, page, and link rows were being restored.
Report a problem from the page where it happened
Signed-in users could report a bug or ask for help from the shared header on Crawl Foundry's main workspace pages. The draft included the current account, workspace, page, and active report area, leaving the user to describe what had happened. The description remained in place if sending failed. Closing a report with unsent text required confirmation, and a successful submission created one report for follow-up. When an error view showed a support code, it offered the same report action and added that code to the draft. The user could describe the interrupted task without copying diagnostic text by hand.
Keyword Research makes interrupted runs recoverable
Keyword Research began distinguishing handoff, waiting, active processing, and finalization instead of showing one indefinite running state. A run that was not accepted in time or stopped responding closed and released its unused balance reservation. An active run could also be cancelled directly. Repeated browser or network requests reused the same research operation instead of creating another reservation. Large imports split into several parts kept one target list and resumed after the last confirmed part. Filtered result pages looked ahead within a strict bound, so later matches did not remain hidden behind an empty page. Customer-facing errors used stable, localized explanations and a support code instead of exposing source or execution details. Authorized operators retained the diagnostics needed for follow-up. Results reused existing keyword and search-result evidence when it was available: cached SERP features and top results, search trends, related parent topics, a clearly labelled traffic estimate, and domain-relative difficulty. Sorting and paginated history remained usable on larger result sets. Missing or partial evidence stayed visibly unknown. Saved research data could be exported as a filtered and sorted CSV. Each result showed its source and, where available, the seed that produced it. A previous run could prefill another request from history. The current price remained visible, and the new request did not start before confirmation.
Domain profiles retain the latest usable evidence
Domain Overview began using a persistent profile. A new analysis extended the existing evidence, while a temporary source failure did not displace an older successful state. Each source state remained traceable as a dated snapshot. A discovered competitor could be carried directly into a new keyword-gap analysis. The competitor and the workspace's own domain were prefilled in their correct roles. The price preview and explicit confirmation still came before the paid request.
Position Tracking keeps market and device contexts separate
Position Tracking began keeping desktop, mobile, country, and language measurements separate in saved results, history, comparisons, and exports. Tracking setup allowed the market and language to be selected instead of using a fixed default without showing it. Report navigation stayed stable while the user switched between reports and tracking jobs. The Cannibalization report remained available from Jobs, and its empty state explained when the recorded search results contained no simultaneous conflict between the site's URLs. The Rankings report added tags plus quick filters for Top 3, winners, losers, and keywords that were not ranking. Filters ran before server pagination, remained in the page URL, and fed the same CSV export. Values that spreadsheet software could interpret as formulas were neutralized in the export. The Positions chart separated improvements, declines, new rankings, losses, and unchanged keywords. Pages showed which landing pages contributed the most estimated traffic. Jobs exposed run coverage and missing data over time. SERP Features distinguished appearances owned by the workspace, owned by a competitor, or lacking reliable attribution. SERP-feature ownership distinguished “not owned” from a result without reliable attribution. Several own URLs counted as cannibalization only when they appeared in the same measured search result. A URL switch between days and older history without concurrency evidence remained a separate change. A multi-device run settled as one group. Impossible progress values were rejected, and stale-run recovery closed abandoned work and released unused reservations. Required search-result updates and ranking history were stored together, preventing a success state when part of that history was missing.
Site Audit applies the 5,000-page limit consistently
Site Audit began applying the 5,000-page limit consistently when a domain, crawl, or recurring audit is configured. Older saved settings with a higher value were reduced for future crawls without changing historical crawl or billing records. Large results also reported their preparation state more precisely. A result did not appear complete while a required summary still lacked eligible pages. Page details loaded evidence for the active section, kept an older crawl selected when opened through a direct link, and explained when that crawl's detailed evidence had expired while its summary remained. URL details stayed inside the same Site Audit navigation and content frame as Pages, with the selected URL in the breadcrumb instead of a second page header. The current product limit remains authoritative for future crawls.
Site Audit comparisons use clear crawl names and matching units
Each Site Audit crawl received a name containing its full date, time, year, and chronological run number for that day. Several crawls from the same day could therefore be selected without relying on their URL or issue counts, and crawls from different years remained distinguishable. Compare kept the earlier and later crawl in chronological order, loaded the available issue groups, and reset the pair when the selected domain changed. Totals distinguished issue occurrences from affected URL instances, so an increase and a decrease used the same unit. Opening a comparison row switched to a crawl in which that issue was present. Compare could load older crawl pages directly, including runs outside the newest 20 entries. It also waited for a run's published results and kept severity changes assigned to the correct category in each crawl.
Site Audit distinguishes a delayed start from a failed crawl
Site Audit began distinguishing a delayed processing start from a failed crawl. An accepted crawl remained queued, retried automatically, and offered a manual retry without creating or charging for a second run. If processing remained delayed, the notice could open a prefilled problem report with the support code and crawl context already attached.
Site Audit separates confirmed defects from review signals
Site Audit began separating confirmed defects from signals that still need review. Corrupted titles and descriptions can be reported as defects, while character length, sparse copy, static render-blocking candidates, and simple cache-policy checks remain advisory unless stronger evidence exists. Intentional forms, galleries, downloads, lightboxes, and decorative `alt=""` markup are not penalized solely for being short or image-led. Repeated anchor and resource patterns are grouped by their shared cause, and samples are labelled as samples. External destinations receive a browser-like confirmation before they are called broken; authentication or challenge pages are not suggested as replacements. Contextual-link advice stays limited to content pages and appears only when the relevant link evidence is complete. Progress reaches 100 percent only after findings, link analysis, Health scoring, and report evidence have been published.
Site Audit keeps issue evidence tied to the selected crawl
The Site Audit Issues tab began anchoring its catalog and history to the selected crawl. Categories, sources, affected-unit types, issue movement, and fixed issues came from published audit data rather than being inferred from the page currently loaded in the browser. Incomplete and failed crawls no longer appeared complete. Bounded counts were labelled as lower bounds, missing history remained visibly missing, and loading more affected URLs or assets retained its loading state instead of briefly showing an empty result. Issue reports kept copying implementation guidance separate from opening an agent-ready handoff. The report retained the full run-level evidence count when only part of the detail list was loaded and explained when more findings could still exist. A failure in one Issues panel stayed inside Site Audit and offered a retry for that section. Reconciliation also closed stale issues when a fresh crawl returned no matching finding, including older cases outside the previous processing window.
Site Audit explains internal link depth and opportunities
Site Audit began calculating click depth as the shortest internal-link path from the start page. Overview and Statistics showed the depth distribution, while Page Explorer could filter and sort URLs by click depth, reachability, incoming links, outgoing links, and internal authority. When two crawls had compatible exact-depth evidence, Overview also showed changes in median depth, P90 depth, and unreachable pages. The Health Score gauge kept a three-digit score legible inside its ring. Internal Links loaded a large site page by page rather than keeping the full URL inventory in the browser. Each URL showed a representative shortest path, named relationship and occurrence counts, and a relative-authority estimate with its basis visible. On desktop, Internal Links and Page Explorer stayed inside the available viewport while long URL and relationship lists scrolled in their own panels. For audits within the 5,000-page limit, internal-link opportunities evaluated the eligible URL set before publishing a bounded action list. Existing links, unsafe targets, incomplete graph evidence, and weak topical matches stayed out. Each action carried separate evidence for target need, topical fit, source authority, placement feasibility, and risk. Ranked actions remained in the Action Center so review and completion used one shared state.
Statistics publish after crawling while page details keep context
Crawl statistics began publishing when crawling finished instead of waiting for later external-link and asset checks. Statistics could therefore show response health, click depth, indexability, and discovery sources while those follow-up checks continued. Opening a page from Pages retained the crawl context. If a copied or shared link did not include a crawl, Crawl Foundry resolved it to the latest crawl in which the page had been recorded instead of rejecting the page until a crawl was selected by hand. The page detail view kept the Site Audit frame. Domain and crawl selectors remained available, and changing the crawl loaded the same page in that crawl. The other Site Audit tabs stayed reachable without using a back link first. Page details stored the selected tab, link direction, filters, asset section, and issue severity in the URL. Previous and next moved through the loaded Pages result without loading a whole crawl into the browser. Inspecting a linked page offered a direct return to its source. Link and issue filters ran before pagination, allowing matches beyond the first loaded batch to appear. A missing page or crawl showed a recoverable explanation instead of a generic failure.
Action Center recommendations now account for confirmed business fit
Action Center began using confirmed business context alongside keyword, ranking, competitor, domain, Search Console, Site Audit, and internal-link evidence. Topics outside the offering were held back, while uncertain matches became questions instead of confident instructions. Difficulty, intent, page readiness, movement, and the site's recorded strength shaped keyword recommendations; missing evidence stayed visible as a data need. The five decision goals now produce different portfolios rather than the same queue in a different order. Recommendations link to their exact working context, can be saved with an editable priority, and include a free deterministic plan with execution and verification steps. If one missing paid data point could change several decisions, Crawl Foundry groups the affected keywords into one proposal and shows scope and price before confirmation. Optional AI suggestions remain separate from confirmed business facts until the user accepts them.
Site Audit waits for crawl evidence before publishing a report
Site Audit began waiting for queued page analysis before marking a crawl complete. The crawl overview, issue groups, URL totals, and Health Score were published from the same completed evidence state rather than from an earlier intermediate result. If final publication was interrupted, Crawl Foundry resumed that step without crawling the website again. Recovery was paced, and temporary write-limit failures received bounded retries. An interrupted publication therefore stayed separate from the completed crawl evidence instead of starting several report rebuilds or rewriting the crawl itself as failed.
Home opens the Position Tracking workspace reliably
Opening Position Tracking from Workspace Home now reached the tracking workspace even when Crawl Foundry still had to prepare part of the view. The setup link stopped at a visible loading state instead of falling into an error screen before the tracking controls appeared. Current access still depends on the workspace's enabled features and permissions.
Site Audit findings now separate evidence from repeated symptoms
Site Audit began recognizing valid alt text, responsive images, accessible link labels, navigation, resource hints, and downloads more precisely. Repeated resource problems are grouped by the affected asset instead of being multiplied across every source page, while the report keeps occurrence counts separate from unique URLs and assets. Overlapping symptoms can share a likely cause without being presented as proven causality. External links now receive browser-compatible confirmation before they are called broken. Login walls and unstable access blocks remain unverifiable, and redirect checks preserve meaningful trailing slashes. Contextual-link findings appear only when the crawl has complete body-link evidence; otherwise the report states that the evidence is partial. A crawl also stays in processing until its findings and Health Score are published, so exports and implementation guidance cannot present an unfinished projection as a completed audit.
Site Audit separates website health from crawl confidence
Site Audit began calculating Health from the pages and checks that were eligible in a particular crawl. Many low-impact findings no longer forced the score onto an artificial floor. Failed or incomplete crawl work appeared in the confidence state instead of being mixed into website health. Every published score was labelled Measured or Provisional. When the available evidence could not support a score, Crawl Foundry showed what was missing instead of filling the gap with a number. The overview also named verified blocker caps, the weakest score categories, and the main drivers as unique affected pages against the real eligible-page denominator. Drivers were not presented as additive point deductions because several rules can overlap on one page. The Sites overview, crawl history, comparisons, portfolio summaries, and Health drop alerts used the same persisted score snapshot. Older summary-only crawls remained visibly unsupported. Incompatible runs produced no Health delta, and alerts compared only measured crawls with matching scoring version and crawl scope.
Site Audit separates live crawl speed from processing time
Running Site Audits began showing their average pages per minute directly below Crawl Time. The rate stopped changing when crawling finished, so later report processing did not make the completed crawl appear slower. Processing Time began with the first processing work, even while crawling was still active. Crawl Time, Processing Time, and Total Time therefore described overlapping phases rather than suggesting that all processing waited for the last fetched page.
Action Center connects evidence across SEO workflows
Action Center began bringing evidence from keyword work, domain analysis, competitors, tracking, connected Search Console coverage, Site Audit, and internal links into one decision guide. Each recommendation showed its source, known limits, prerequisite, responsible product area, and a way to check the result. Business fit and ranking feasibility could hold back an attractive keyword instead of turning demand alone into a task. A workspace with little history received a conservative setup sequence, while a workspace with no qualifying action received an honest empty state rather than filler. Users could save recommendations to **My actions**, change their priority and status, and keep that list across later refreshes. Optional deeper data or an AI plan started only after confirmation. The AI plan worked on one selected action, used a reduced SEO evidence packet, excluded private documents and credentials, and could not silently replace the deterministic guide after a failed or invalid response. Historical plans remained visible while a refresh ran or failed.
Dashboard and Activity now explain every signal in context
Help on Home began following the workspace briefing from identity and data state through keywords, competitors, technical health, priorities, recent activity, product readiness, and learning resources. Each step named the source of the summary and linked to the detailed documentation for the workflow that owned it. Activity explained its bounded 28-day summary, newest-first event history, counters, status meanings, notification filters, support evidence, and access to older history batches. The guide also stated the ownership boundary: Activity recorded outcomes, while live controls and detailed state remained in the product area that created the operation.
Crawl Foundry turns SEO data into a guided next step
This release connected Domain Overview, competitors, Action Center, Keyword Research, the Keyword Database, and the historical Position Tracking surface into one guided workflow. Workspace Home showed which evidence existed, what was ready for review, and, where tracking was configured, when the next scheduled ranking check was planned. A step counted as complete only when the corresponding result existed in the workspace. Domain and competitor views opened with a concise situation report: what the current data indicated, why it might matter, which evidence supported the statement, and where to continue. Position Tracking separated movements marked for immediate review from later checks and watch items. Missing search volume did not create an artificial urgency signal. Keyword Research added selection recipes for ideas labeled reachable, commercial, or supported by competitor evidence. When previously purchased top-10 results existed for the same market, language, and device context, the workspace could group keywords by observed SERP overlap without another data purchase. Site Audit grouped findings by likely remedy area and labeled that group as a planning aid rather than a proven shared cause. Action Center kept the immediate queue focused while the broader list grew or shrank with qualifying evidence. Each admitted action stated its intended benefit, estimated effort, evidence, prerequisites, and next step. The workflow described here is dated; current navigation and tracking availability remain authoritative in the product.
Workspace Home expands its evidence briefing
This release expanded Workspace Home beyond Site Audit and the Keyword Database. The page was organized as a briefing on the domain, possible next actions, recent outcomes, and the evidence held by each product area. The workspace domain led the page. Its summary band showed observed organic keywords, top-10 rankings and their change since the previous analysis, estimated traffic value, open Site Audit findings, and keyword-inventory size. Each number linked to the product area that owned its evidence. Below it, one column drew possible work from the existing priority list. A second, **What happened**, listed completed crawls, domain analyses, keyword-research runs, and data refreshes across products, with a link back to each source. Site Audit and the Keyword Database retained their larger widgets, including list selection and finding trends. Compact cards represented Domain Overview, Keyword Research, Tracking, and Competitors. A product area with data showed its headline value and last update. An area without setup explained what evidence it could provide and offered one starting route instead of displaying zero as if it were a measurement. Areas unavailable to the current role were omitted. The final row linked to documentation and the latest Blog and Glossary entries available at the time of the page request. The layout used the available width on large screens and stacked the same zones on smaller screens. Brain, Research, and Content cards were removed from Home while those areas were still being developed, although their historical pages remained in navigation at this date. This entry does not describe the current Workspace Home or current feature availability.
Research help now follows each decision step
The Help walkthroughs in Keyword Research and Competitive Research began following the workbench from top to bottom instead of ending after a short introduction. They covered research modes, seed terms and domains, market context, live estimates, run evidence, filters, term groups, candidate selection, imports, saved runs, domain comparisons, history, market position, competitive signals, keyword gaps, coverage, list actions, and competitor details. Enter moved to the next visible step. Each step also linked to its localized Help Center topic when the reader needed the full explanation. In Domain Overview, the run selector stayed in the domain context row and remained readable in both light and dark themes. This entry records the guidance and page structure introduced in this release; the current workbench remains the source of truth for available modes and controls.
Keyword Research now starts from consistent local search signals
Keyword Research began generating suggestion ideas for the market selected in the research form rather than the location from which processing happened to run. The suggestions served as additional starting points and still passed through the same relevance, deduplication, and metrics checks before entering the result. Repeated Quick Research runs therefore used the intended market context without shrinking the candidate pool. The change improved consistency; it did not make separate live suggestion runs identical.
Workspace pages now share one clear page header
Home, Action Center, Domain, Research, Competitors, Tracking, Briefs, Site Audit, Activity, and Keywords adopted the same compact orientation bar. Page title, relevant workspace context, and primary page actions stayed in a predictable place. Controls that belonged to a particular view or dataset remained directly below it. In the Keyword Database, the main metrics shared a single-line context rail above the view controls, leaving more room for the keyword table without hiding the current data context. Header actions also followed one visual rule. Help used a blue information treatment, while ordinary page actions used the primary button color. Home kept the header focused on Help instead of duplicating Keyword Database and Site Audit navigation. Help opened a guided walkthrough for the current page and followed its visible areas in order. Domain Overview covered snapshot context, authority, rankings, competitors, keywords, links, and growth. The historical Tracking surface adapted the guide to Overview, Positions, Pages, SERP features, competitors, cannibalization, and schedules. Site Audit covered its ten workbench views, while a crawled URL adapted Help to Overview, findings, links, assets, schema, changes, or technical evidence. Each step linked to the matching localized documentation topic. Keywords kept its more detailed database onboarding behind the same entry point. This entry records the workspace navigation available at the time, including surfaces that are no longer part of the current core product.
Position Tracking stays responsive as your history grows
Position Tracking began isolating failures to the affected view. If one data source was temporarily unavailable, the tracker selector, other tabs, and schedule management remained usable. Only the affected area showed its support code and retry action. An unavailable optional Site Audit link also stopped without hiding the recorded ranking-page data. History loads were bounded to the selected context instead of sending every measurement to the browser. Trends showed the latest 30 points, competitor analysis loaded only while its view was open, the active competitor set was capped at 20, and schedule lists showed the latest 30 entries with a notice when older entries existed. Before a schedule could start, its cost estimate had to be current. Invalid keyword, language, location, device, or budget input was rejected consistently. List-based trackers supported up to 500 keywords per run at this release, while a larger list received a visible limit response instead of starting an operation that could not complete. The overview retained all four headline-metric positions when a value was unavailable and explained the missing data instead of substituting zero. The creation workbench was spacious but bounded, keeping domain, keyword selection, competitors, and schedule settings in view without covering the entire application. The Orders view used the available workspace width and kept one creation action in the page header. These tracking limits and surfaces describe the product at this date, not the current Crawl Foundry feature set.
Site Audit: one place for crawl history and one place to inspect a run
The History tab became **Crawls** and separated run history from inspection. Its table listed each stored crawl with crawled URL count, health, open findings, and the change in each value from the previous run. Selecting a crawl turned the same tab into a run inspector. The inspector summarized the recorded outcome in plain language. While a crawl was active, "Happening now" showed the current fetch stage and queue composition, and the inspector opened that section by default. The stored crawl log used pagination rather than a small sample. Users could page through its URL records, filter findings or redirects, and open a page's detail view. Error messages appeared below the URL as readable sentences rather than internal codes. Tables scrolled inside their own area and used the available height even when only one run existed. A URL detail page added a breadcrumb and back action consistent with the rest of the workspace. **Monitoring** was renamed **Automation** because schedules could start audits rather than only observe them. Each schedule showed its previous run and outcome and linked to that crawl. Schedules and alerts could be edited as well as paused or deleted, and each used its own creation form. **Crawl now** started an audit immediately. Schedule and alert cards retained their empty state after all records were deleted, so the tab still explained what could be configured.
Domain Overview is safer, richer and easier to reuse
Domain analyses gained a paginated history, visible storage limits, and a confirmed delete action. Repeating the same request after a network interruption reused its operation identity instead of creating another analysis or balance reservation. A paid request did not start when the number or value of active reservations could not be established within the release limits. Organic keyword rows added the observed ranking page, estimated keyword traffic, and competition when supplied by the source. The competitor panel retained its compact comparison and added a table with shared keywords, total observed organic keywords, average position, and estimated traffic. A selection dialog connected the snapshot to the rest of the workspace. Selected rows could be stored in a deduplicated `Domain · example.com` list in the Keyword Database while retaining available volume, difficulty, CPC, competition, intent, and language data. The same selection could seed Keyword Research, with the 100-seed limit shown before navigation. Across the paid Research, Tracking, Brief, Import, and Site Audit actions available at the time, Crawl Foundry stopped before an external data request when the active-reservation total could not be calculated exactly. The current product remains authoritative for today's paid actions, limits, and retention.
Site Audit: full site structure, link trends and cleaner statistics
The Site structure table began showing every folder recorded in the selected crawl instead of hiding smaller groups. It used the available screen height, scrolled independently, and kept column headers visible. The Internal Links overview added up to six per-crawl trends: internal links, average links per page, orphan pages, external links, `nofollow`, and links to redirects. Link Explorer and the link map also used the available viewport height. The run overview kept the same statistics and source values but separated them into flatter sections. Each finding's "Why this happens" section linked to a relevant glossary entry, including canonical tags, hreflang, and redirect chains. Domain and run pickers closed when focus moved elsewhere. Asset findings displayed the complete asset and source-page URLs and opened their evidence directly. The Issues overview reused the compact agent actions from an individual finding. **Agent prompt** opened a Markdown preview for the findings in the current filtered view. **Agent endpoint** created a scoped pull address only after explicit confirmation, without expanding a separate export workspace above the list. An optional guide explained the handoff to coding agents, from repository access through validation and a confirming crawl. The endpoint panel explained when a pull workflow was more useful than a copied prompt, how to pass the authenticated request to a coding agent, and why the temporary credential had to remain private. The credential was no longer placed in the URL, expired after one hour, and could be revoked from the same panel. Cancelled and failed crawls received a plain-language state and a possible next step, while internal error codes stayed out of the customer-facing view.
Site Audit monitoring makes schedules and delivery states visible
The Monitoring tab reduced initial setup to cadence and recipients. Weekly was the recommended default at this release, while advanced controls exposed timing, budget, URL limits, and triggers. Each recurring schedule showed its next planned run and could be paused, resumed, or deleted. If a run was skipped, for example because the prepaid balance was insufficient, the schedule card displayed the recorded reason. Notification cards sat beside the schedule, and an activity feed listed each triggered event with the known email-delivery state. The status distinguished recorded processing states; it did not treat acceptance by the email service as proof that the recipient had read the message. These scheduling defaults describe the release at this date.
Site Audit corrects statistics loading and click-depth scope
This release corrected two Statistics-tab behaviors. Statistics for a finished crawl became available without leaving the tab on "Statistics are being prepared" after processing had completed. Click depth was also limited to pages reached through recorded internal links from the start page. A page discovered only through the sitemap no longer increased the internally reachable group and appeared as "Not linked" for this measure. Existing crawl results kept their stored values; the corrected distribution began with the next crawl.
Keyword Research now explores topics more intelligently
Quick Research began expanding each topic through closely related searches and up to three related angles. It tested questions, prepositions, modifiers, comparisons, year and local intent, and A-Z variants, then sent a bounded candidate set to one shared metrics check. Deep Research extended that exploration through adaptive rounds. It followed relevant new angles, stopped when the additional yield weakened, and enriched a bounded set before returning up to 2,000 deduplicated keywords under the release limits. Starting keywords were counted while the user typed. The live estimate appeared in a fixed-width start button, preventing price refreshes from moving the controls or repeating the reserved amount elsewhere. Quick Research validated at most 360 suggestion candidates and returned at most 600 deduplicated keywords. At this release, the confirmed estimate was the charge ceiling, so a higher upstream cost was not added after completion. Before a paid data request, the run verified that it was still attached to the correct workspace and dataset. A failed handoff released its reservation. The activity view distinguished delivery, configuration, network, timeout, quota, and data-source failures instead of collapsing them into one message. These methods, limits, and price rules describe the product at this date; the current checkout remains authoritative.
Competitor Workbench: compare the right domains at the right depth
Competitor Workbench began turning domains observed in tracked search results into a maintained catalog. Users could add, remove, or pin relevant competitors while keeping the changing SERP landscape separate from the curated comparison set. At this release, up to 20 competitors could be active at once, and the larger candidate catalog remained paginated. A quick comparison covered the main authority, traffic, keyword, and link signals. Deep analysis added keyword gaps, intent, difficulty, top pages, and more detailed link evidence. Before either run, checkout itemized the estimate and reused recent snapshots or completed modules, so only missing data was requested. Results stayed as dated snapshots: changing the active set prepared a new comparison and price preview instead of rewriting an earlier result.
Internal Links rebuilt around traceable relationships
This release rebuilt the Internal Links area around the relationships and occurrences recorded by a Site Audit crawl. The **Overview** showed how many internal and external links the crawl found, how many pointed to redirects or carried `nofollow`, which crawled pages had no recorded incoming internal link, and which pages received the most internal links. Each count and row opened the corresponding filtered evidence rather than ending at a summary. In the **Explorer**, selecting a page opened its title, complete URL, navigation breadcrumb, and counts for incoming, outgoing, and external links. Full-width tabs separated incoming, outgoing, external, and backlink relationships, while placement filters distinguished content, navigation, footer, and other recorded regions. A **List / Map** switch changed the presentation of the same selection. **Page details** opened the complete page report and its findings without losing the selected link context. Repeated links between the same source and target appeared as one relationship. Its row retained the occurrence count, distinct anchor texts, and recorded page regions such as navigation, header, and footer, so deduplication did not remove the evidence. The **All links** table used its internal, external, broken, redirect, and `nofollow` summary tiles as filters. Any row could open the Explorer at its source page. All counts remained bounded by the selected crawl and any separately labeled backlink source.
Keyword Research moves from a form to a workbench
This release moved Keyword Research from a single form into a results workbench. A run could begin with keywords or competitor domains, a market, and a research depth. The checkout showed its estimate before confirmation, and the result opened as a filterable table with the available search volume, CPC, difficulty, and intent for each row. Match modes separated All, Broad, Phrase, Exact, Related, and Questions. A term rail showed frequent words in the current result set, while filters covered volume, difficulty, intent, source, and included or excluded terms. These controls worked on the data already returned by the run and did not start another paid request. A marker identified keywords already stored in the Keyword Database, and an "only new" filter hid them. Import remained explicit: users selected the rows to keep and chose an existing or new list. Recent searches stored each run with its market, keyword count, and settled cost under the retention rules available at the time. Reopening and filtering a stored result did not purchase the same data again. This entry records the July 2026 workflow; current retention, research methods, and prices remain authoritative in the product.
Position Tracking: competitor discovery, SERP-feature landscape, and a visual overview
The Competitors view began listing domains observed repeatedly in the top 10 for the tracked keyword set, including domains the user had not added manually. Each row showed shared keywords, best observed position, and average position. A domain could be added to or removed from the comparison. When retained search results covered earlier measurements, a newly selected competitor could use that stored history instead of starting with an empty series. The SERP-features view summarized which recorded results contained AI Overviews, featured snippets, local packs, image packs, and other features, how many tracked keywords were affected, and where the tracked domain was observed in that feature. Selecting a feature opened the corresponding keyword set. The overview gave visibility, estimated traffic, average position, and share of voice separate cards with trend charts. A stacked chart showed how the top 3, 10, 20, and 100 counts changed between runs. A fact line stated the tracked domain, device, language, market, keyword and competitor counts, search depth, and last update. Missing search volume remained unknown in estimated-traffic views instead of being rendered as zero. The Competitors view could compare visibility, estimated traffic, average position, top-10 count, or share of voice across the selected domains. The keyword table used pagination, and the overview trend could also plot how many tracked keywords had any recorded ranking. The creation wizard allowed the tracked domain to be changed in its first step. In the Positions view, each keyword had its own series and each completed measurement added a point. These tracking surfaces describe the product at this date and are not a promise about Crawl Foundry's current feature set.
Clearer paths to every public Crawl Foundry page
This release aligned public routes across the platform, features, solutions, learning resources, comparisons, and company information. Existing customer and comparison addresses stayed stable where possible, while retired routes redirected to the closest maintained destination rather than ending at a dead page. Navigation and page directories exposed the same route groups in each supported language. Public listings also tied their update signal to a meaningful source change, so an unchanged page did not receive a new date merely because the site was rebuilt. The release improved discovery; it did not certify that every linked page or translation was complete.
Site Audit: a real page for every URL
The Pages view began with internal URLs, while a visible scope control switched between Internal, External, and All. Selecting a page opened a dedicated URL route instead of an overlay. It showed the crawl-derived score, findings, HTTP status, response time, word count, and link counts, with separate tabs for findings, links and the recorded crawl path, assets, structured data, changes since the previous crawl, and technical facts. Following a linked URL opened that URL's page, so normal browser history worked across the investigation. Existing deep links redirected to the new route rather than breaking. The overview grouped its readiness checks in one list. Selecting a severity count opened the matching findings, and the crawl path displayed the recorded internal-link route as a numbered chain. A **Copy for agent** action prepared a prompt with URL context, captured evidence, and the proposed fix for one finding or the current set of open findings. The prompt remained a handoff for review, not proof that the suggested change was correct for every codebase.
New About, Security, and Contact pages
This release introduced dedicated About, Security, and Contact pages. The About page used an interactive, scroll-led story to explain how Crawl Foundry connects search evidence, prioritization, and later review. The Security page documented the normal sign-in and workspace access boundaries, stated the limits of those claims, and provided a route for reporting a security concern. The Contact page added a working topic-based form that delivered messages to the team without promising a fixed response time. The pages have continued to evolve since this release, so their current copy remains the authoritative description.
Domain Overview brings key evidence into one snapshot
Domain Overview began combining authority, backlinks, referring domains, estimated organic traffic, ranking keywords, competitors, top pages, anchors, and highly linked pages in one requested snapshot. Before the run, checkout showed the selected market, keyword depth, and estimated price; the final charge stayed within that confirmed estimate. Available history could show monthly traffic and keyword movement, while older analyses remained reachable through pagination. Users could send selected keywords to the Keyword Database or Keyword Research, start a separate analysis from an observed competitor, or compare several domains side by side. A comparison kept authority, traffic, keywords, links, history, gaps, and top pages together and reused recent data where the release contract allowed it. These were dated snapshots and deliberate handoffs, not an automatic import of every observed result.
Workspace home puts current signals before inventory
The redesigned workspace home opened with the signals available for the website: Site Health, average position and its trend, keyword-inventory size, and the number of recorded opportunities. Each summary linked to its underlying view, so a headline number could be checked rather than treated as a complete verdict. The "What to do now" section translated available findings into linked prompts such as reviewing high-severity issues, inspecting a keyword that had dropped, or deciding how to handle recorded opportunities. These were prioritized starting points, not proof that every suggested action would improve performance. Site Audit showed its health indicator, error and warning counts, and frequent findings. Keywords showed position distribution, trend, and the largest movements for a selected list. The Brain and Research widgets summarized the knowledge base and recent reports that belonged to the historical product at this date. Unused tools explained what would appear in their empty area, while a new workspace offered two starting points instead of charts without data. This entry records that dated dashboard, not the current product navigation.
Keyword discovery: pick a depth, not a data source
This release replaced four source-oriented packages with three task-oriented research depths. Quick Research returned related keywords and ideas with their included metrics through the lower-cost route. Deep Research searched more broadly and combined observed Google search suggestions with a metrics check. Competitor Analysis continued to return keywords found for the selected competitor domains. Crawl Foundry selected the required sources and removed duplicate candidates before presenting the results. Before each run, the checkout showed the current estimate and available balance. Settlement used the data actually delivered and was bounded by the confirmed estimate. Method cards calculated their estimate from the current seeds, an empty workspace could start from the keywords observed for its own domain, and a "Questions only" filter isolated question-shaped candidates. The Keyword Database also added a direct lookup path. A search for a term not yet stored could create the keyword and open its detail panel; paid metrics appeared only after the user confirmed the data request. These research tiers and prices describe the product at this date. The current checkout remains the authority for today's methods, coverage, and cost.
Optimizer presets now stay in sync
The historical Optimizer began saving the exact strategy and custom focus selected by the user, including rapid consecutive changes. Existing custom presets remained readable, while an invalid stored value returned to a supported default instead of reopening a project with an unsupported strategy or malformed weights. Result warnings were also localized in English, German, French, and Spanish. This entry describes the Writer and Optimizer available at the time, not a current Crawl Foundry workflow.
Position Tracking: one surface for rankings, pages, and competitors
This release expanded Position Tracking from a single chart into a dedicated workspace. The overview showed visibility, estimated traffic, average position, the number of tracked keywords in the top 3, 10, 20, and 100, and which terms had entered or left those ranges since the previous measurement. A trend chart covered the selected measurement window. Dedicated views separated the evidence: a position table with movement, search volume, CPC, and the observed ranking URL; a Pages view grouping tracked terms by URL; a Tags comparison for keyword groups; a SERP-features view for recorded featured snippets and AI Overview citations; a competitor comparison using the same keyword set; and a cannibalization report for terms whose observed ranking URL changed between two pages. A context switcher changed device or location across the workspace, while schedule management remained in its own tab. A user could start a schedule by selecting keywords in the Keyword Database and choosing **Track**, without first creating a list. These recurring tracking features describe the product at this date; they are not a promise about Crawl Foundry's current feature set.
Research runs now recover more safely
The historical Live Research workflow began validating progress, discoveries, cost, completion, and error updates before they reached the report. An incomplete response ended in an explicit error state instead of leaving malformed report data on the page. A late completion from a cancelled run was also prevented from stopping or overwriting a newer session. The fix reduced a specific stale-update risk in the Research product available at the time; it did not guarantee recovery from every interruption.
A clearer way into the Crawl Foundry knowledge system
This release rebuilt the Resources overview and Learning Center as separate entry points for different reading goals. Resources explained the role of the Learning Center, Blog, SEO Glossary, documentation, and Changelog, how they related to one another, and included the blog analyses published at the time. The Learning Center grouped four paths into Keywords and Technical SEO. Each path showed its stated difficulty, estimated duration, and workflow steps, while a goal selector suggested a place to begin. Guides displayed their position in the track and linked to the previous and next path. These were navigation aids, not a claim that the resource set was complete or that one suggested path fit every reader.
A clearer Site Audit and a calmer workspace
This release organized Site Audit around three questions: What did the crawl say about the site's health? Which findings deserve review first? Where did they occur? The overview combined a health score and trend, clickable finding counts, a "Fix first" list, and category tiles that opened the corresponding filtered findings. Sites, Overview, Issues, Pages, Links, Compare, History, Monitoring, and Settings moved into one tab row instead of a second sidebar. Each view opened the latest finished crawl by default. Selecting an older run in the run picker kept a visible route back to the latest result. The Compare tab showed differences between two crawls: findings no longer observed, new findings, and pages added or removed. It documented change between measurements without claiming that a particular fix caused the difference. The application sidebar could collapse to an icon rail with the **[** shortcut. The prepaid balance appeared above the profile, and organization switching moved into the account menu. These navigation details record the interface at this date.
See your site's structure at a glance
The Pages view added a **Site structure** lens beside the flat URL inventory. Its folder tree represented the URLs recorded in the selected crawl. Each folder showed its page count, its share of the crawled site, and how many recorded pages were broken or marked `noindex`. Folders also displayed the open findings for their pages and highlighted those containing critical or high-severity findings. They expanded in place; URL, finding, broken-page, redirect, and `noindex` columns were sortable; and search narrowed the tree without flattening its hierarchy. A folder at any depth opened Page Explorer with a "path starts with" filter already applied. The Pages toolbar kept search and one Filters button visible, with the complete filter workbench and column manager behind that control.
SEO work starts with the task now
This release separated the main SEO jobs in the workspace navigation: Keyword Research for demand discovery, Competitive Research for competitor evidence, Position Tracking for the selected keyword sets monitored at the time, Site Audit for crawl-based technical findings, and the Action Center for prioritization. Keywords remained the shared inventory for lists, filters, Matrix, Kanban, Activity, paid enrichment, and bulk actions. Existing bookmarks continued to resolve to the relevant workflow, while the sidebar and internal links opened each dedicated area directly. The navigation changed without splitting the underlying records. Research results could still be saved to Keywords, tracked lists retained their identity, and recommendations opened the view that owned the evidence or action. The compact sidebar allowed enough room for translated tool labels and kept the complete destination name available to assistive technology. Internal-link opportunities also received one source of record. Site Audit stored the source, target, anchor, and lifecycle evidence in Internal Links. The Action Center could surface prioritized recommendations, but opened the matching filtered Site Audit workbench for the underlying detail. This entry records the product structure at the time, including the historical Position Tracking surface.
A more focused Site Audit workspace
This release narrowed the first Site Audit workspace to crawl setup, health, history, findings, pages, links, monitoring, and settings. Overlapping or unfinished areas were removed from the navigation instead of appearing as if they were ready. Link Explorer remained the dedicated place for site structure and link-path evidence, while the other views stayed focused on the crawl results they actually owned.
Workspace limits explain upgrades and recovery
Workspace creation began explaining directly in the form when the current plan's allowance had been reached. The message showed current usage and the available limit, preserved the user's input, and linked to the relevant plan options without exposing internal error details. Subscription checkout also recognized returning customers whose accounts predated the organization billing model, avoiding a duplicate-customer failure during an upgrade. If a later plan supported fewer workspaces, the release introduced a 30-day adjustment period instead of immediately removing or hiding existing work. Teams could choose which workspaces stayed active and retained read, export, billing, and recovery access for the others. A deliberately deleted workspace remained restorable for 30 days; permanent deletion followed only after the recovery window and a verified cleanup. Organization administrators could also create a private portable export before making the final decision. These were the recovery rules introduced at the time, not a substitute for checking the current plan and retention terms.
Action Center prioritizes the next SEO moves
The Action Center was reorganized around one decision flow: understand the current SEO position, choose the immediate goal, compare prioritized opportunities, and open the specialist view responsible for the selected work. Teams could focus the ranking on balanced priorities, visibility, quick wins, ranking protection, or technical improvements. Each recommendation showed its rationale, intended outcome, estimated effort, and available evidence before the user followed it. The prioritization ranked the evidence Crawl Foundry had; it did not guarantee that the proposed action would produce the expected result. On smaller screens, the same decision brief opened in a focused mobile workspace instead of dropping the supporting context.
Keyword intelligence and Site Audit, connected
This release connected the Keyword Database and Site Audit as two parts of the same SEO workflow. A team could move from a search opportunity to the technical findings for an important page, inspect the affected URLs, and compare later crawl evidence with an earlier run. The comparison showed what changed between measurements; it did not prove that one particular intervention caused the change.
Crawl-wide link paths, with their evidence
Link Explorer began using the complete URL inventory recorded for a crawl instead of a small sample loaded into the current view. Selecting a page showed the linking pages and each recorded anchor occurrence. Incoming links, internal and external outgoing links, and self-links stayed separate, with anchor placement and DOM position available as evidence. Pages with many connections opened in a compact cluster view with separate relationship and occurrence totals. A cluster could be expanded into its concrete URLs, while the detailed relationship view remained available. External inbound links were labeled as backlink data from a separate source rather than being inferred from the website crawl.
A clearer command center for SEO work
The workspace dashboard was refocused on the Keyword Database and Site Audit. It brought keyword coverage, search intent, available site-health signals, crawled URL counts, and current priorities into one summary. Each item led to the responsible product area, so the dashboard remained an entry point to the underlying evidence rather than pretending to contain the full diagnosis.
A clearer SEO command center
This companion dashboard update brought Keyword Database and Site Audit summaries into one responsive overview. Key totals, available ranking signals, technical-issue trends, gains, losses, and current priorities could be scanned before opening the underlying workspace. The overview also distinguished current data from unavailable history. An empty chart was labeled as missing evidence instead of implying progress that had not been measured.
Site Audit monitoring alerts are easier to trust
Site Audit monitoring began with simpler recommended presets, while advanced scheduling controls stayed out of the way until someone needed them. Alert history also stopped treating acceptance by the email service as proof of delivery. It distinguished confirmed delivery, bounce, complaint, retry, and a digest still waiting in the queue, so teams could see what was known about a notification instead of assuming it had reached the recipient.
Brief-led content generation
The historical Writer could now begin with the user's brief instead of forcing research to be the first step. It showed whether project knowledge was active and gave explicit directions about exclusions, tone, format, and topic focus priority over generic SEO suggestions. Structure, drafting, and optimization therefore followed the same instruction set from the start. A trust panel made the limits of a draft more visible. It showed when support was missing, evidence was weak, a risky claim had been checked or adjusted, or the Optimizer had drifted from the evidence and user direction. Recurring writing instructions could be corrected, pinned, demoted, or forgotten, while the current project brief could override a saved rule. Before review and scoring, the workflow could remove an unsupported high-risk statement, soften an unsupported absolute claim, or reword it cautiously when the available project knowledge justified that change. Precision edits, Optimizer rewrites, and manual editor suggestions followed the same boundary. They were expected to respect explicit exclusions and avoid introducing concrete claims that the available project knowledge did not support. Quality problems from real runs could be summarized without copying private article text or project knowledge into a report. The safe QA bundle covered weak evidence, unsupported claims, Optimizer drift, ignored directions, and thin topic coverage, and a snapshot could be kept in workspace history. A RAG QA view in the Brain area added a review queue, recurring-pattern summaries, Golden Case drafts, knowledge-connection views, documented browser checks, and calibration packages. A proposed quality change still required explicit review before it could affect live behavior. This entry records the Writer and Brain workflow available at the time; those surfaces are not part of Crawl Foundry's current core product promise.
Site Audit foundation for workspace crawls
This release introduced the first workspace flow for a technical Site Audit. A team could add a website, configure the crawl, review the expected use of its prepaid balance, and then start the run. The resulting workspace brought the URL inventory, grouped findings, individual URL evidence, internal links, and the site tree together without presenting the initial foundation as a complete diagnosis of every technical SEO problem.
Keyword discovery, tracking, and reliability
This release connected discovery, stored SERP history, competitor snapshots, score signals, activity history, recurring position checks, and clearer cost feedback inside the Keyword Database. A keyword could move from an initial idea into the monitored set without being copied between separate tools, while paid data actions exposed their cost context more clearly. Recurring tracking described here was part of the product at this date; it is not a promise about the current Crawl Foundry feature set.
Keyword workbench visual refresh
The Keyword Database was reorganized for large lists. Metrics, table columns, row actions, and bulk-selection states followed a clearer visual order, so users could compare opportunities without repeatedly losing their place. This release changed the workbench presentation and interaction states; it did not introduce a new keyword metric.
Editorial publishing workflows
Publishing checks for the Blog, glossary, and public service notices became more explicit. Editors could see missing localization, preview an entry, and confirm its publishing state before it reached a customer-facing page. The release did not remove editorial review; it made incomplete translations and readiness problems easier to catch before publication.
Writer, research, and Brain stability
The Writer, Research, and Digital Brain became less prone to state collisions during longer sessions. Saving, restoring, AI-assisted edits, research sessions, and Brain panels handled newer work more carefully when users moved between planning and writing. The update reduced the risk of restoring stale context or overwriting a more recent edit; it did not promise that work could never be lost. These were parts of the historical Writer and Brain workflow available at the time.
Team, billing, and review controls
This release tightened the boundaries between team administration, organization billing, invitations, client access, and editorial review. Account and access changes were handled more consistently, while content and research work moved through a more explicit approval cycle. Teams could tell more easily what was waiting for review and which access or billing context applied before they acted.
Public site and help center refresh
This release repaired inconsistencies across the public website, Help Center, glossary, sitemap, and localized pages. Navigation labels, metadata, translations, and mobile layouts followed the same conventions more consistently, making public routes easier to find and use. It was a presentation and discoverability update, not a claim that every article or translation was complete.
SEO audit and research foundations
This release introduced Crawl Foundry's first on-page audit workspace and made partial research data easier to recognize. Reports separated source health, URL-level findings, and missing evidence, so an incomplete dataset was not presented with the same confidence as a well-supported result. Teams could see where the available evidence supported a decision and where another source or a fresh check was still needed.
Dashboard and product controls
Workspace dashboards were reorganized to show more useful status without spreading it across as many screens. Billing state and feature access became easier to distinguish, while the help center was reorganized around onboarding, day-to-day workflows, team use, and troubleshooting. The release changed how existing information was presented; it did not add a new measurement or automated recommendation.
Keyword Database workbench
The Keyword Database brought imports, tags, clustering, requested SERP evidence, detail views, and bulk actions into one workspace. Teams could organize a keyword set before buying additional data, inspect the available competitive evidence, and choose which terms were worth a paid refresh instead of enriching the entire list by default.
Research intelligence and keyword opportunities
The Research view began explaining search intent, topical fit, and missing coverage alongside expanded keyword ideas. Opportunities were grouped into practical labels such as "easy win", "competitive", and "needs more work" so teams could decide what deserved a closer review. The labels were prioritization aids, not promises about ranking or the effort a page would require.
Content guidance and brand voice
The writing workflow could now reuse a stronger brand-voice profile, more relevant internal-link suggestions, and the defaults stored for its workspace. Teams no longer had to restate the same style, quality, and linking instructions whenever they opened a new content project. These were features of the Writer available at the time, not a description of Crawl Foundry's current workflow.
Workspace defaults and knowledge intake
New projects could now inherit their target score, author profile, language, and expected word count from the workspace. The website intake used by the Digital Brain was also made more reliable, giving the Writer a more consistent source for brand voice and content guidance without asking teams to re-enter the same settings for every project. These were features of the workflow available at the time.
Team workspaces and EUR billing
This release introduced organizations, roles, invitations, project-level access, and shared billing. Agencies and in-house teams could separate brands or client work inside one workspace structure. Usage was shown in EUR rather than an internal credit unit, making purchase decisions easier to compare with actual spend.
Brain, scoring, and optimization upgrades
The Digital Brain could now connect related facts and favor stronger sources when supplying context to the Writer. Optimization also became section-specific: a weak passage could be revised with the evidence behind that section instead of sending the entire article through a generic rewrite. These functions were part of the Writer and Brain workflow available at the time; this entry does not describe the current product.
Research and content quality foundation
This release introduced the first shared foundation for research-backed content work. The research workflow could bring together signals from search results, community discussions, academic sources, video, and public web pages. A companion score separated relevance, factual support, readability, and basic SEO checks, so teams could see why a draft needed work instead of receiving one unexplained number.
Product interface foundation
This release established the visual system used by the early Crawl Foundry interface: higher-contrast surfaces, consistent spacing, shared components, and patterns that also worked in dark mode. Dense SEO data became easier to scan, and moving between projects and the writing workspace involved fewer visual changes.